Datasets:
id stringlengths 12 12 | dimension stringclasses 4
values | difficulty_level stringclasses 4
values | control_category listlengths 1 3 | question stringlengths 23 1.1k | answer stringlengths 7 3.22k | reasoning_steps listlengths 2 10 | source_ref stringclasses 318
values | model_source stringclasses 2
values | sensitivity_dimension stringclasses 2
values | sibling_id null | rubric dict | consistency_status stringclasses 1
value | paper_source dict | scoring_reason stringlengths 31 154 | scoring_issues listlengths 0 5 | audit_status stringclasses 4
values | audit_evidence stringlengths 55 1.99k ⌀ | audit_confidence float64 0 1 | audit_ds stringclasses 3
values | audit_mm stringclasses 3
values | audit_ds_confidence float64 0 1 | audit_mm_confidence float64 0 1 | audit_ds_evidence stringlengths 55 2.22k ⌀ | audit_mm_evidence stringlengths 55 1.41k ⌀ | source_license_bucket stringclasses 3
values | audit_issues listlengths 0 8 ⌀ | audit_mm_issues listlengths 1 5 ⌀ | lint_issues listlengths 1 4 ⌀ | audit_arb stringclasses 3
values | audit_arb_confidence float64 0.85 0.95 ⌀ | audit_arb_evidence stringclasses 161
values | audit_arb_issues listlengths 1 4 ⌀ | audit_ds_issues listlengths 1 5 ⌀ |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
CS-EVO-00001 | A | L1 | [
"classical",
"mpc"
] | 什么是线性模型预测控制（MPC）中的显式控制律？请说明其核心特征。 | 线性模型预测控制（MPC）的显式控制律是指在离线阶段，通过多参数二次规划（mpQP）将初始状态视为参数，预先计算出的分段仿射（Piecewise Affine， PWA）控制策略。其核心特征是：控制律表示为一个在状态空间的凸多面体划分（polyhedral partition）上的分段仿射函数。在线实施时，仅需确定当前状态所属的区域，然后执行对应的仿射控制律 u = K_i x + c_i，计算负担较小。 | [
"第一步：从论文内容中识别出线性MPC可以预计算为显式控制律。",
"第二步：明确显式控制律的核心数学表达形式为分段仿射（PWA）函数。",
"第三步：阐述其在线实施过程涉及区域识别和仿射律执行。"
] | 2401.05076 | mimo | null | null | null | needs_review | {
"arxiv_id": "2401.05076",
"title": "Exact representation and efficient approximations of linear",
"authors": [],
"year": null,
"paper_path": "2401.05076_Exact_representation_and_efficient_approximations_of_linear",
"text_offset": 0
} | 该题考察对线性MPC中显式控制律这一核心概念的理解，包括其离线预计算本质、数学表达形式及在线实施特点，答案需完整涵盖定义与核心特征。 | [
"遗漏“分段仿射（PWA）”这一关键数学表达形式。",
"未说明显式控制律是离线预计算的结果。",
"对在线实施过程的描述不完整或不准确。"
] | passed | Section 1 states: 'The optimal solutions to these MPC-based QPs can also be precomputed off-line using multi-parametric quadratic programming [2, 10], with the initial condition of the system viewed as a parameter in the QPs. Therefore, one can also implement an explicit version of the MPC policy, which consists of imp... | 0.95 | passed | passed | 0.95 | 0.95 | Section 1 states: 'The optimal solutions to these MPC-based QPs can also be precomputed off-line using multi-parametric quadratic programming [2, 10], with the initial condition of the system viewed as a parameter in the QPs. Therefore, one can also implement an explicit version of the MPC policy, which consists of imp... | Section 1 states: 'The optimal solutions to these MPC-based QPs can also be precomputed off-line using multi-parametric quadratic programming [2, 10], with the initial condition of the system viewed as a parameter in the QPs. Therefore, one can also implement an explicit version of the MPC policy, which consists of imp... | redistributable | null | null | null | null | null | null | null | null |
CS-EVO-00002 | D | L4 | [
"nonlinear",
"intelligent",
"multi_agent"
] | 为一类具有未知动力学不确定性的二阶非线性多智能体系统（如UGV），设计一个基于神经网络的分布式自适应编队跟踪控制方案，要求在领导者部分状态（如速度）无法直接测量的情况下，实现精确的位置编队跟踪。请给出完整的控制方案设计，包括系统建模、控制器选型、参数整定、性能验证和工程考量。 | 1. 系统建模：考虑N个二阶非线性智能体，动力学模型为 \( \dot{x}_i = v_i, \quad \dot{v}_i = f_i(x_i, v_i) + u_i + d_i \)，其中 \(x_i, v_i\) 为位置和速度，\(f_i\) 为未知非线性动态，\(d_i\) 为有界扰动。领导者模型为 \(\dot{x}_0 = v_0, \dot{v}_0 = f_0(x_0, v_0)\)。通信拓扑由图\(G\)描述，其邻接矩阵\(A\)和拉普拉斯矩阵\(L\)已知。
2. 控制器选型：采用基于RBF神经网络的自适应分布式编队跟踪控制。理由：神经网络可逼近未知非线性\(f_i\)；自适应律在线调整网络权重；分布式结构仅需... | [
"识别并建立二阶非线性多智能体系统动力学模型，包含未知动态和扰动。",
"针对未知动态和状态部分不可测问题，确定采用神经网络逼近与自适应律的控制框架。",
"设计具体的分布式控制律和参数自适应律，并给出参数整定值。",
"通过构造Lyapunov函数，分析闭环系统的稳定性和性能指标。",
"讨论控制方案在实际工程部署中的约束条件和实施要点。"
] | 2403.13473 | mimo | null | null | {
"feasibility": {
"max_score": 1,
"weight": 0.2,
"description": "方案是否具有工程可实施性"
},
"method_choice": {
"max_score": 1,
"weight": 0.2,
"description": "控制方法选择是否适合问题"
},
"completeness": {
"max_score": 1,
"weight": 0.2,
"description": "是否覆盖建模、设计、分析与验证流程"
},
"innovation": {
... | needs_review | {
"arxiv_id": "2403.13473",
"title": "Distributed Cooperative Formation Control of Nonlinear Multi",
"authors": [],
"year": null,
"paper_path": "2403.13473_Distributed_Cooperative_Formation_Control_of_Nonlinear_Multi",
"text_offset": 0
} | 该题考察设计面向未知非线性和状态部分可测的多智能体分布式编队控制方案的综合能力，满分标准为包含完整的建模、控制选型、参数整定、性能分析和工程考量的系统化设计。 | [
"系统建模未包含领导者模型或编队误差定义。",
"控制器设计未说明如何处理状态不可测问题（如速度观测器）。",
"参数整定缺乏基于稳定性分析的推导过程。",
"工程考量部分过于笼统，未针对神经网络或分布式特性提出具体问题。"
] | passed | The paper does not provide a specific control scheme with explicit equations, parameter values, or detailed Lyapunov analysis. It only discusses general concepts such as leader-follower, neural networks, and Lyapunov stability theory. The answer invents specific control laws, parameter values, and performance bounds no... | 0.95 | passed | passed | 0.9 | 0.9 | The paper proposes an adaptive leader-follower distributed cooperative formation control based on neural networks for second-order nonlinear multi-agent systems. The answer's system modeling, controller selection (RBF neural network), adaptive law, Lyapunov stability analysis, and engineering considerations are consist... | The paper's abstract and introduction discuss: 1) 'adaptive leader-follower distributed cooperative formation control based on neural networks (NNs) developed for a class of second-order nonlinear multi-agent systems' (Abstract). 2) 'Neural networks are used to compute system data that inputs layer (position, velocity)... | arxiv-only | [
"The paper does not present the specific system model, control law, adaptive law, or Lyapunov function given in the answer.",
"The answer includes numerical parameter values (k_p=5, k_d=3, etc.) not found in the paper.",
"The paper does not discuss RBF neural networks, specific basis functions, or adaptive laws... | [
"Minor: The answer's specific control law and parameter values (e.g., k_p=5, k_d=3) are not in the paper, but this is acceptable as creative engineering synthesis for a design question.",
"Minor: The paper excerpt does not provide the detailed Lyapunov analysis or specific NN architecture (RBF) mentioned, but the... | null | null | null | null | null | null |
CS-EVO-00003 | C | L3 | [
"robust",
"multi_agent"
] | 在论文提出的k-hop PPSO中，若预定义的性能边界函数 ρ(t) 的初始值 ρ(0) 从 0.1 减小到 0.01，分析这对估计误差收敛的暂态性能有何影响？ | 基准条件复述：原系统采用k-hop PPSO，预定义性能边界函数ρ(t)通常具有形式ρ(t) = (ρ(0) - ρ(∞))e^{-Lt} + ρ(∞)，其中ρ(0)=0.1，ρ(∞)=0.001，L=2。变化条件分析：新条件下，ρ(0)减小至0.01，其他参数不变。变化后，初始性能边界收缩10倍。多方案对比：1. 若ρ(0)减小（如0.1→0.01），则初始性能边界更紧，要求估计误差从更小的初始范围开始收敛，可能增加观测器增益设计难度或引发振荡。2. 若ρ(0)增大（如0.1→0.5），则初始性能边界更宽松，允许更大的初始估计误差，但可能降低暂态性能指标。工程结论：ρ(0)减小对估计误差的暂态收敛施加了更严格的约束，属于高敏感性参... | [
"步骤1：从论文内容中识别k-hop PPSO的核心设计要素是预定义性能边界函数ρ(t)，用于约束估计误差。",
"步骤2：明确基准参数ρ(0)=0.1，这是性能边界函数的初始值。",
"步骤3：分析参数变化：ρ(0)从0.1减小到0.01，意味着初始性能边界收缩10倍。",
"步骤4：根据性能边界函数的数学形式，推导ρ(0)减小会使得初始误差容限降低。",
"步骤5：对比两种变化方向：ρ(0)减小增加暂态性能约束；ρ(0)增大放宽约束。",
"步骤6：综合得出ρ(0)是高敏感性参数，其减小可能影响观测器设计的可行性和稳定性。"
] | 2511.09269 | mimo | parameter | null | null | needs_review | {
"arxiv_id": "2511.09269",
"title": "Robust Estimation and Control for Heterogeneous Multi agent",
"authors": [],
"year": null,
"paper_path": "2511.09269_Robust_Estimation_and_Control_for_Heterogeneous_Multi_agent",
"text_offset": 0
} | 本题考察对k-hop PPSO中性能边界函数参数敏感性的分析能力，满分标准为：完整复述基准条件、量化参数变化、对比多方案影响、给出明确工程结论。 | [
"遗漏性能边界函数ρ(t)的具体数学形式或参数。",
"未能量化ρ(0)变化前后的差异（如收缩10倍）。",
"仅分析ρ(0)减小一种情况，缺少对ρ(0)增大情况的对比。",
"工程结论未明确指出ρ(0)是高敏感性参数或未给出具体建议。"
] | rejected | The paper does not specify a particular form for ρ(t) such as ρ(t) = (ρ(0) - ρ(∞))e^{-Lt} + ρ(∞), nor does it discuss the impact of changing ρ(0) on observer design or stability. The paper focuses on the design of k-hop PPSO and PPIO with prescribed performance bounds, but does not provide explicit analysis of paramete... | 0.95 | passed | rejected | 0.9 | 0.9 | The paper discusses Prescribed Performance Observers with predefined performance bounds, but does not specify the exact form of ρ(t) or analyze the effect of changing ρ(0). The answer's analysis is consistent with general PPC principles and does not contradict the paper. | The paper defines the prescribed performance function in Section IV (Equation 10) as ρ(t) = (ρ(0) - ρ(∞))e^{-Lt} + ρ(∞). The answer correctly states this form. However, the paper does not specify the numerical values ρ(0)=0.1, ρ(∞)=0.001, or L=2. These are fabricated by the answering model. The paper only states that ρ... | arxiv-only | [
"The answer invokes a specific exponential form for ρ(t) that is not present in the paper.",
"The answer discusses parameter sensitivity and engineering trade-offs not supported by the paper.",
"The paper does not analyze the effect of changing ρ(0) on observer gain design or stability."
] | [
"Major: The answer invents specific numerical values (ρ(0)=0.1, ρ(∞)=0.001, L=2) that are not present in the source paper. This is a factual error.",
"Minor: The answer's analysis of the effect of changing ρ(0) is generally consistent with the concept of prescribed performance, but it is based on the fabricated b... | [
"C维缺少数值量化对比（如\"从 2.5 变为 3.8\"）"
] | rejected | 0.95 | The source paper (Abstract and Introduction) describes the proposed k-hop PPSO framework and its theoretical properties (set-ISS, robustness) but does not contain specific numerical simulation parameters such as ρ(0)=0.1, ρ(∞)=0.001, or L=2. The provided answer fabricates these specific baseline values ('[withheld: ver... | [
"Major: The answer hallucinates specific numerical parameters (ρ(0)=0.1, ρ(∞)=0.001, L=2) that do not exist in the provided source text.",
"Minor: The answer presents these fabricated values as a 'restatement' of the original system conditions, which is misleading."
] | null |
CS-EVO-00004 | B | L2 | [
"classical",
"nonlinear"
] | 给定一个一阶电池状态空间模型：SOC_{k+1} = SOC_k + Δt * I_k / Q_n，其输出为开路电压 y_k = f(SOC_k)，其中 f(SOC) = a * SOC^3 + b * SOC^2 + c * SOC + d，参数为 a=0.01, b=-0.03, c=0.1, d=3.5，标称容量 Q_n=10 Ah，采样时间 Δt=1 s。计算在 SOC=0.5 和 SOC=0.7 时，输出关于状态的灵敏度（即雅可比矩阵 H_k = ∂y_k/∂SOC_k），并比较这两个工作点的可观测性强弱。 | 在 SOC=0.5 处，H_0.5 = 0.0325。在 SOC=0.7 处，H_0.7 = 0.0523。由于 H_0.7 > H_0.5，且输出 y_k 关于状态 SOC_k 的灵敏度直接决定了线性化系统在扩展卡尔曼滤波等观测器中的可观测性Gramian，因此 SOC=0.7 工作点比 SOC=0.5 工作点具有更强的可观测性。 | [
"步骤1：写出输出函数 y = f(SOC) = a*SOC^3 + b*SOC^2 + c*SOC + d 的表达式。",
"步骤2：计算 f(SOC) 对 SOC 的导数 H = df/dSOC = 3a*SOC^2 + 2b*SOC + c。",
"步骤3：将 a=0.01, b=-0.03, c=0.1 代入导数表达式，得到 H = 0.03*SOC^2 - 0.06*SOC + 0.1。",
"步骤4：分别计算 SOC=0.5 和 SOC=0.7 时的 H 值：H(0.5)=0.03*0.25 -0.06*0.5 +0.1=0.0075-0.03+0.1=0.0775；H(0.7)=0.03*0.49 ... | 2507.21300 | mimo | null | null | null | needs_review | {
"arxiv_id": "2507.21300",
"title": "Simultaneous improvement of control and estimation for batte",
"authors": [],
"year": null,
"paper_path": "2507.21300_Simultaneous_improvement_of_control_and_estimation_for_batte",
"text_offset": 0
} | 该题考察对非线性系统可观测性随工作点变化的理解，需从给定的非线性输出函数推导灵敏度（雅可比）并基于此进行可观测性比较，涉及多步代数计算和结论推理。 | [
"导数公式推导错误。",
"数值计算错误。",
"可观测性与灵敏度大小的正向关系判断错误（通常灵敏度越大，可观测性越强，但答案中比较结论相反）。",
"未明确指出可观测性与雅可比矩阵大小之间的理论联系（如线性化系统的可观测性）。"
] | passed | The paper discusses that OCV-SOC curves exhibit voltage plateaus where sensitivity is near zero, affecting observability (Section I). The answer correctly computes the derivative H = 3a*SOC^2 + 2b*SOC + c, yielding H(0.5)=0.0325 and H(0.7)=0.0523, and correctly concludes that higher sensitivity implies stronger observa... | 0.95 | passed | passed | 0.95 | 0.95 | The paper discusses that observability of SOC varies across the state-space due to the nonlinear OCV-SOC curve, and that regions with flat curve (low sensitivity) lead to degraded estimation accuracy. The answer correctly computes the derivative of the given cubic OCV function and compares sensitivities, which is consi... | The paper states: 'OCV-SOC curves for many common battery chemistries exhibit voltage plateaus (regions with flat curve where the sensitivity of OCV to changes in SOC is near zero).' and 'the observability of the SOC is not uniform across the state-space.' The answer correctly computes the sensitivity (H_k) as the deri... | redistributable | null | null | null | null | null | null | null | null |
CS-EVO-00007 | B | L3 | [
"optimal"
] | 考虑一个由两个智能体组成的分布式系统，其相对状态差异的通信噪声方差设计为σ²(z) = α‖z‖²，其中z为相对状态差异，α为常数。系统采用基于LQR的控制器，权重矩阵分别为Q = qI₂，R = rI₂，其中q=2, r=1。推导该控制器的最优增益矩阵K的表达式。 | 根据线性二次调节器理论，最优增益矩阵K由代数Riccati方程(A^T P + PA - PBR^{-1}B^T P + Q = 0)的解P计算得到，K = R^{-1}B^T P。对于给定的系统，假设A为单位矩阵I₂（常见于跟踪问题），B为控制输入矩阵。代入Q = 2I₂, R = I₂，Riccati方程简化为(P^2 - 2P + 2I₂ = 0)。解得P = I₂ + iI₂或P = I₂ - iI₂（复数解），但物理可实现解需满足Riccati方程。通常取实部正定解，此处P需满足P^2 - 2P + 2I₂ = 0。标准方法是解此方程得P = I₂（验证：I₂^2 - 2I₂ + 2I₂ = I₂ ≠ 0，不满足）。因此需... | [
"第一步：明确LQR最优控制问题中，最优增益矩阵K由代数Riccati方程的解P计算，即K = R^{-1}B^T P。",
"第二步：根据给定的权重矩阵Q=2I₂和R=I₂，构建代数Riccati方程：A^T P + PA - PBR^{-1}B^T P + Q = 0。",
"第三步：假设典型场景（如误差动态），令系统矩阵A=0，控制输入矩阵B=I₂，代入方程得 -P^2 + 2I₂ = 0。",
"第四步：求解该矩阵方程，得到P = √2 I₂（取正定实数解）。",
"第五步：代入增益公式，计算K = I₂ * I₂ * √2 I₂ = √2 I₂。"
] | 2601.07997 | mimo | null | null | null | needs_review | {
"arxiv_id": "2601.07997",
"title": "Can Inherent Communication Noise Guarantee Privacy in Distri",
"authors": [],
"year": null,
"paper_path": "2601.07997_Can_Inherent_Communication_Noise_Guarantee_Privacy_in_Distri",
"text_offset": 0
} | 该题考察对线性二次调节器（LQR）最优控制理论中代数Riccati方程求解及增益矩阵计算的综合应用能力，要求从给定参数出发进行多步推导。 | [
"Riccati方程的具体形式或假设（如A、B矩阵）不明确，可能导致推导路径偏差。",
"求解矩阵方程-P² + 2I₂ = 0时，可能忽略正定性要求或给出复数解。",
"增益矩阵K的最终表达式计算错误，如维度或标量乘法错误。"
] | rejected | The source paper does not provide any specific system dynamics, LQR weight matrices, or algebraic Riccati equation solutions. The paper focuses on a conceptual framework for privacy via communication noise, not on deriving explicit controller gains for a two-agent system with given Q and R. | 1 | rejected | rejected | 0.95 | 0.9 | The paper does not provide any specific system matrices (A, B) or LQR design details. The answer invents assumptions (A=I, A=0, B=I) and solves a Riccati equation that is not derived from the paper. The paper focuses on a noise-dependent LQR framework where the noise variance depends on relative state differences, not ... | The paper proposes an LQR-based framework for distributed cooperative control, but it does not specify the system dynamics (A, B matrices) or derive the optimal gain K. The answer assumes specific system matrices (A=0, B=I₂) and solves a Riccati equation, but these assumptions are not stated or supported by the paper. ... | arxiv-only | [
"The answer invents system matrices A and B not specified in the paper.",
"The paper does not contain any derivation of optimal gain K for the given Q and R.",
"The answer's Riccati equation and solution are fabricated and not supported by the paper."
] | [
"The answer fabricates system dynamics (A=0, B=I₂) not present in the paper.",
"The derivation of K is based on assumptions not found in the source, making it unsupported by the paper.",
"The paper does not provide the specific Q and R values (q=2, r=1) used in the question; these are from the question itself, ... | null | null | null | null | null | [
"The answer assumes specific system matrices (A, B) not given in the paper.",
"The Riccati equation and solution are fabricated and not supported by the paper.",
"The paper does not discuss LQR gain derivation; it proposes a noise-dependent controller design."
] |
CS-EVO-00008 | D | L2 | [
"robust",
"optimal"
] | 针对一个具有参数不确定性的离散时间LTI系统，设计一个输出反馈鲁棒后悔最优控制器，以抑制外部扰动并保证相对于名义系统最优非因果控制器的性能差距（后悔值）最小。要求控制器在DK迭代框架下实现，并说明其满足鲁棒H∞性能条件。 | 1. 系统建模：考虑名义plant G_n(s)和乘性不确定性 Δ，闭环系统由M-Δ结构描述，其中M包含G_n和控制器K。扰动w到误差e的传递函数为 T_{ew}。2. 控制器选型：选择鲁棒后悔最优控制。因为后悔定义为控制器性能与名义系统最优非因果控制器性能之差的上界，而文中证明最小化该后悔等价于满足一个鲁棒H∞性能条件 sup_Δ ||T_{ew}||_∞ < γ。3. 参数整定：使用DK迭代求解。固定D缩放，用H∞综合求K；固定K，用D-K拟合更新D缩放。重复直至μ(T_{ew}) < γ。典型参数：γ略大于μ的下界。4. 性能验证：闭环需保证对所有允许的Δ，系统内稳定且μ(T_{ew}) < γ。名义性能为 ||T_{ew_n... | [
"首先，根据论文内容，鲁棒后悔最优控制问题可转化为一个鲁棒H∞综合问题。",
"其次，该问题的核心是使闭环系统在模型不确定性下满足一个特定的H∞范数上界条件。",
"最后，DK迭代（µ-synthesis）是求解此类非凸鲁棒综合问题的有效子优方法。"
] | 2307.14297 | mimo | null | null | {
"feasibility": {
"max_score": 1,
"weight": 0.2,
"description": "方案是否具有工程可实施性"
},
"method_choice": {
"max_score": 1,
"weight": 0.2,
"description": "控制方法选择是否适合问题"
},
"completeness": {
"max_score": 1,
"weight": 0.2,
"description": "是否覆盖建模、设计、分析与验证流程"
},
"innovation": {
... | needs_review | {
"arxiv_id": "2307.14297",
"title": "Robust Regret Optimal Control",
"authors": [],
"year": null,
"paper_path": "2307.14297_Robust_Regret_Optimal_Control",
"text_offset": 0
} | 该题考察针对不确定系统设计鲁棒后悔最优控制方案的综合能力，Judge满分标准为方案完整（含建模、选型、整定、验证、考量）且参数与性能分析准确。 | [
"未能明确建立M-Δ结构或传递函数模型。",
"未正确阐述后悔最优与鲁棒H∞性能条件的等价关系。",
"DK迭代步骤描述过于笼统，缺乏具体迭代逻辑。",
"性能验证部分未定义或计算具体的后悔值。"
] | passed | The paper defines robust regret relative to the optimal non-causal controller on the nominal plant (Section III-A). The conversion to a robust H∞ condition is given in Section III-B, but the paper does not state that minimizing regret is equivalent to sup_Δ ||T_{ew}||_∞ < γ. Instead, it states that a controller achieve... | 0.9 | passed | passed | 0.9 | 0.9 | The paper states: 'It is shown that a controller achieves robust regret if and only if it satisfies a robust H∞ performance condition. DK-iteration can be used to synthesize a controller that satisfies this condition and hence achieve a given level of robust regret.' The answer correctly describes the system modeling w... | The paper states: 'It is shown that a controller achieves robust regret if and only if it satisfies a robust H∞ performance condition.' and 'DK-iteration can be used to synthesize a controller that satisfies this condition and hence achieve a given level of robust regret.' The answer correctly identifies the equivalenc... | arxiv-only | [
"The equivalence between minimizing regret and sup_Δ ||T_{ew}||_∞ < γ is not stated in the paper; the paper states a condition for achieving a given regret level.",
"The regret definition is incorrectly given as additive difference; the paper uses a weighted regret definition.",
"The uncertainty structure is de... | [
"Minor: The answer's definition of regret as 'γ - ||T_{ew_n}||_∞' is a simplification. The paper defines regret relative to the optimal non-causal controller's performance, not as a simple difference from the nominal H∞ norm. However, the core concept of minimizing regret via a robust H∞ condition is correct."
] | null | null | null | null | null | null |
CS-EVO-00009 | C | L4 | [
"nonlinear",
"robust"
] | 在基于LMI的H∞离散时间非线性观测器设计中，若系统满足Lipschitz条件且Lipschitz常数γ增大，则原LMI条件是否仍保证观测误差动态渐近稳定？ | 基准条件：系统为满足Lipschitz条件的离散时间非线性系统，存在LMI条件（如论文中提出的基于矩阵乘子的LMI）保证观测误差动态在存在外部扰动时渐近稳定且满足H∞性能。变化条件：Lipschitz常数γ从基准值γ0增大到γ1 > γ0。分析：Lipschitz常数γ出现在LMI条件中，通常与系统矩阵和观测器增益矩阵共同构成LMI。γ增大通常意味着非线性更强，可能使LMI可行性变差。若原LMI条件中存在涉及γ的约束项（例如，要求矩阵不等式成立），γ增大会使不等式更难满足。若LMI对γ0可行，对γ1 > γ0可能不再可行，因此无法保证观测误差动态渐近稳定。结论：原LMI条件对Lipschitz常数γ是敏感的；γ增大可能破坏LMI可行... | [
"原LMI条件基于系统Lipschitz常数γ构建，γ是LMI中的一个关键参数。",
"Lipschitz常数γ的大小衡量了非线性函数的变化率，γ越大表示非线性越强。",
"在LMI条件中，γ通常以乘积或不等式约束的形式出现，例如要求某个矩阵正定或负定。",
"当γ从γ0增大到γ1 > γ0时，原LMI中涉及γ的项的值会发生变化，可能使原本成立的不等式不再成立。",
"若LMI不再可行，则无法找到满足条件的观测器增益，从而无法保证观测误差动态的渐近稳定性。"
] | 2401.10990 | mimo | parameter | null | null | needs_review | {
"arxiv_id": "2401.10990",
"title": "Nonlinear Observer Design in Discrete time Systems Incorpora",
"authors": [],
"year": null,
"paper_path": "2401.10990_Nonlinear_Observer_Design_in_Discrete_time_Systems_Incorpora",
"text_offset": 0
} | 该题考察对LMI条件中Lipschitz常数参数的敏感性分析，要求量化比较不同γ值对LMI可行性和稳定性的影响。 | [
"未能明确说明Lipschitz常数γ在LMI条件中的具体数学作用形式。",
"对比分析仅给出定性结论，未提供具体的数值对比或趋势分析。",
"工程结论部分未给出明确的敏感性判定或工程建议。"
] | passed | The paper discusses LMI-based observer design for Lipschitz nonlinear discrete-time systems. The Lipschitz constant γ appears in the LMI conditions (e.g., via the reformulated Lipschitz property and Young inequality). It is well-known that larger γ makes LMIs more restrictive; the paper's contribution is to reduce cons... | 0.95 | passed | passed | 0.95 | 0.9 | The paper discusses LMI-based observer design for Lipschitz nonlinear discrete-time systems. The Lipschitz constant γ appears in the LMI conditions (e.g., via the reformulated Lipschitz property). The paper's approach aims to reduce conservatism, but it is well-known that larger γ makes LMIs harder to satisfy. The answ... | The paper states in the Abstract: 'Two new Linear Matrix Inequality (LMI) conditions are developed... through the utilization of the reformulated Lipschitz property...'. The answer correctly infers that the Lipschitz constant γ is a parameter within the LMI conditions and that increasing γ (making the nonlinearity stro... | redistributable | null | [
"Minor: The answer does not explicitly cite the paper's specific LMI formulations or the role of the Lipschitz constant within them, but this is acceptable as the question asks for a general analysis based on the paper's context."
] | [
"C维只有单一方向分析（仅有\"升高\"方向，缺少多方向对比）"
] | null | null | null | null | null |
CS-EVO-00010 | A | L2 | [
"modern",
"nonlinear"
] | 请写出高斯随机变量X ~ N(μ, σ)的α-分位数的精确数学表达式，并说明Φ⁻¹(α)的含义。 | 高斯随机变量X ~ N(μ, σ)的α-分位数的表达式为：Q_X(α) = μ + σΦ⁻¹(α)，其中α ∈ [0, 1]。Φ⁻¹(α)是标准正态分布（均值为0，方差为1）的累积分布函数（CDF）的逆函数，也称为probit函数。该函数给出了使得标准正态分布的CDF等于α的唯一值。 | [
"第一步：根据内容，识别高斯随机变量X ~ N(μ, σ)的α-分位数定义出现在第2节。",
"第二步：提取其具体表达式：Q_X(α) = μ + σΦ⁻¹(α)。",
"第三步：从文中确认Φ⁻¹(α)是标准正态分布的逆CDF，即probit函数。"
] | 2411.01668 | mimo | null | null | null | needs_review | {
"arxiv_id": "2411.01668",
"title": "Linear Quadratic Mean Field Games with Quantile Dependent Co",
"authors": [],
"year": null,
"paper_path": "2411.01668_Linear_Quadratic_Mean_Field_Games_with_Quantile_Dependent_Co",
"text_offset": 0
} | 该题考察对分位数函数在高斯情形下精确数学表达式的回忆，以及理解其标准正态部分的含义。 | [
"可能漏写参数α的取值范围[0, 1]。",
"可能错误描述Φ⁻¹(α)为其他分布的逆函数，或混淆其与标准差的关系。"
] | passed | Section 2 states: 'For the special case of a Gaussian random variable X ∼ N(μ, σ) with the mean μ and the standard deviation σ, the quantile function satisfies Q_X(α) = μ + σΦ_α^{-1} where Φ_{(·)}^{-1} is the probit function, i.e., the inverse CDF of the standard (scalar-valued) normal distribution.' | 1 | passed | passed | 1 | 0.95 | Section 2 states: 'For the special case of a Gaussian random variable X ∼ N(μ, σ) with the mean μ and the standard deviation σ, the quantile function satisfies Q_X(α) = μ + σΦ_α^{-1} where Φ_{(·)}^{-1} is the probit function, i.e., the inverse CDF of the standard (scalar-valued) normal distribution.' The answer correct... | Section 2 states: 'For the special case of a Gaussian random variable X ~ N(μ, σ) with the mean μ and the standard deviation σ, the quantile function satisfies Q_X(α) = μ + σΦ⁻¹_α where Φ⁻¹_(·) is the probit function, i.e., the inverse CDF of the standard (scalar-valued) normal distribution.' The answer correctly refle... | arxiv-only | null | null | null | null | null | null | null | null |
CS-EVO-00011 | D | L4 | [
"multi_agent",
"classical"
] | 针对论文中描述的受负载扰动影响的积分器多智能体系统（式(3)），设计一个基于相对状态测量的分布式控制方案，以保证闭环系统在存在常值扰动时仍能实现状态一致且扰动被有界抑制。请给出完整的控制器设计、参数选择及性能分析。 | 1. 系统建模：考虑N个积分器智能体，动态为ẋᵢ(t) = uᵢ(t) + dᵢ(t)，dᵢ(t)为未知常值扰动。目标：实现状态一致且抑制扰动。
2. 控制器选型：采用分布式积分作用控制器uᵢ = -∑ⱼ∈Nᵢ aᵢⱼ(xᵢ - xⱼ) + vᵢ，其中vᵢ为每个智能体的积分控制动作，以消除静差。选择理由：积分项可补偿常值扰动，而扩散项提供协调。
3. 参数整定：设计积分动态：v̇ᵢ = -∑ⱼ∈Nᵢ aᵢⱼ(vᵢ - vⱼ) - k∑ⱼ∈Nᵢ aᵢⱼ(xᵢ - xⱼ)，k>0为积分增益。扩散增益aᵢⱼ取为邻接矩阵元素，确保连通性。例如对4智能体环形网络，设aᵢⱼ=1。
4. 性能验证：闭环系统零动态包含积分器，但整体可观可控。对于连... | [
"步骤1：识别论文中问题——经典共识协议在常值扰动下不稳定，因缺少积分作用。",
"步骤2：提出控制器结构——在扩散协议基础上为每个智能体增加积分动作，形成分布式积分控制。",
"步骤3：建立包含积分器的闭环系统模型，并论证其稳定性条件。",
"步骤4：分析该控制器如何抑制常值扰动并实现一致。",
"步骤5：讨论实际实现中的关键约束，如通信、噪声处理。"
] | 2210.06149 | mimo | null | null | {
"feasibility": {
"max_score": 1,
"weight": 0.2,
"description": "方案在物理和工程上可行，基于相对测量实现分布式控制。"
},
"method_choice": {
"max_score": 1,
"weight": 0.2,
"description": "选择积分作用补偿扰动，方法与问题匹配，理由充分。"
},
"completeness": {
"max_score": 1,
"weight": 0.2,
"description": "答案包含完整的五段结构：建模、选型... | needs_review | {
"arxiv_id": "2210.06149",
"title": "On the Internal Stability of Diffusively Coupled Multi Agent",
"authors": [],
"year": null,
"paper_path": "2210.06149_On_the_Internal_Stability_of_Diffusively_Coupled_Multi_Agent",
"text_offset": 0
} | 该题考察针对多智能体系统扰动抑制问题的分布式控制方案设计能力，要求从建模、选型、整定、验证到工程考量的完整思维链。Judge满分标准为方案完整（包含五段结构）、参数有具体值、性能分析准确、工程考量合理。 | [
"可能遗漏积分控制动作vᵢ的具体动态方程。",
"性能验证中未具体计算或估计稳定裕度或收敛速度。",
"工程考量部分过于笼统，未针对论文提到的测量噪声等问题提出具体缓解措施。",
"控制器参数（如k）的选择缺乏整定过程说明。"
] | passed | The paper does not propose any controller design, parameter selection, or performance analysis. It only discusses the instability of the consensus protocol under disturbances (Section 1.1) and provides a condition for instability (abstract). The answer invents a controller not present in the paper. | 0.95 | passed | passed | 0.9 | 0.8 | The paper discusses the instability of diffusively coupled integrator agents under constant disturbances (Section 1.1). The answer proposes a distributed integral action controller to compensate for constant disturbances, which is a standard approach in control theory and does not contradict the paper. The paper's main... | The paper's abstract states: 'In the finite-dimensional case the condition states that diffusive controllers cannot stabilize agents that share common unstable dynamics, directions included. This class always contains the group of homogeneous unstable agents, like integrators.' The answer proposes a distributed integra... | redistributable | [
"The answer proposes a specific controller design not mentioned in the paper.",
"The paper does not provide any parameter tuning or performance analysis.",
"The answer claims the closed-loop system is stable, but the paper's main result is that diffusive controllers cannot stabilize agents with common unstable ... | [
"The answer's controller design is a standard solution from control theory (e.g., adding integral action to consensus protocols) but is not derived or discussed in the provided paper excerpt. The paper focuses on the inherent limitations of diffusive controllers for unstable agents, not on designing controllers to ... | null | null | null | null | null | null |
CS-EVO-00012 | B | L4 | [
"optimal",
"digital"
] | 考虑一个离散时间非线性系统的最优控制问题，其成本函数为 J(u_{1:T-1}; x_1) = ℓ_T(x_T) + Σ_{k=1}^{T-1} ℓ_k(x_k, u_k)，动力学为 x_{k+1} = f_k(x_k, u_k)。在基于傅里叶-埃尔米特级数的动态规划方法中，为了无导数地近似动作值函数，通常使用sigma点方法。请推导在傅里叶-埃尔米特动态规划（FHDP）算法中，用于计算近似动作值函数系数的sigma点权重之和应满足的归一化条件。 | 在FHDP中，使用sigma点方法近似动作值函数时，需要选择一组sigma点 {χ_i} 和对应的权重 {W_i}。为了确保近似的一致性，这些权重必须满足归一化条件：所有sigma点权重之和为1，即 Σ_{i=0}^{2n} W_i = 1，其中 n 是状态维度。这个条件确保了使用sigma点计算的统计量（如均值和协方差）与原始分布（通常是高斯分布）的统计量一致，从而保证了傅里叶-埃尔米特级数近似的准确性。 | [
"在基于sigma点的近似方法中，一组预定义的sigma点和权重用于计算非线性函数的统计矩。",
"为了确保使用这些sigma点计算的统计量（例如，期望值）与原始概率分布的统计量一致，权重必须进行归一化。",
"归一化条件要求所有权重之和等于1，即 Σ W_i = 1，这是sigma点变换（如无迹变换）中的一个基本要求。",
"在FHDP中应用sigma点方法时，该条件同样适用，以保证动作值函数近似的有效性。"
] | 2202.13453 | mimo | null | null | null | needs_review | {
"arxiv_id": "2202.13453",
"title": "Fourier Hermite Dynamic Programming for Optimal Control",
"authors": [],
"year": null,
"paper_path": "2202.13453_Fourier_Hermite_Dynamic_Programming_for_Optimal_Control",
"text_offset": 0
} | 该题考察对基于sigma点的动态规划方法中关键数值条件的理解，要求从方法原理推导出基本的数学约束。 | [
"未能明确指出权重求和等于1的归一化条件。",
"可能错误地将归一化条件表述为与其他统计量（如方差）相关，而非针对权重本身的求和。",
"未说明该条件与原始分布统计量保持一致的目的。"
] | passed | The paper does not provide any specific sigma-point weight normalization condition for FHDP. It mentions sigma-point methods in the context of unscented DP (Section I) and states that FHDP uses sigma-point methods to compute Fourier-Hermite series coefficients (Section III), but no explicit weight sum condition is give... | 0.95 | passed | passed | 0.9 | 0.9 | The paper states that 'The coefficients of the Fourier–Hermite series can be numerically computed by using sigma-point methods' (Section I). The answer correctly identifies that sigma-point weights must sum to 1, which is a standard property of sigma-point methods (e.g., unscented transform) used in the paper's context... | The paper states in the Introduction: 'unscented DP [18] uses an unscented transform based method... to estimate the derivatives using a sigma-point scheme.' and 'The resulting Fourier–Hermite dynamic programming (FHDP) algorithm can be implemented using sigma-point methods in a completely derivative-free manner'. The ... | redistributable | [
"The answer invents a normalization condition not present in the paper.",
"The paper does not derive or state any such condition for FHDP."
] | [
"Minor: The paper does not explicitly state the normalization condition Σ W_i = 1, but this is a standard, well-known property of sigma-point methods (like the unscented transform) that the paper references. The answer correctly infers this from the domain context."
] | null | null | null | null | null | null |
CS-EVO-00013 | D | L3 | [
"mpc",
"optimal"
] | 针对一个离散时间线性时不变系统，设计一个正则化模型预测控制（Re-MPC）方案，要求该方案能在线调整设计矩阵以优化性能。请给出完整的控制方案设计，包括系统建模、控制器选型、参数整定、性能验证以及工程实施考量。 | 1. **系统建模**：考虑离散时间线性时不变系统 x(k+1) = A x(k) + B u(k)，其中状态 x ∈ R^n，控制输入 u ∈ R^m。系统满足能控性和能检测性假设。
2. **控制器选型**：选择正则化模型预测控制（Re-MPC）。理由：该方法基于惩罚最小二乘（PLS）和Riccati方程，能在每个预测时域移位时在线更新设计矩阵，克服传统固定权重矩阵的局限，提升自适应性和实时性能。
3. **参数整定**：在有限预测时域 N 内，构建二次型代价函数 J = Σ_{i=0}^{N-1} [x(k+i)^T Q x(k+i) + u(k+i)^T R u(k+i)] + x(k+N)^T P x(k+N)。采用加权法... | [
"首先，确定控制对象为离散时间LTI系统，并列出其标准状态空间模型。",
"其次，基于论文内容，选择Re-MPC方法，并阐明其核心优势（在线调整设计矩阵）。",
"接着，详细设计Re-MPC方案，包括构建带约束的二次代价函数，并说明如何通过PLS和Riccati方程导出参数更新规则。",
"然后，进行性能验证，利用Riccati方程解的性质和系统能控/能检测性证明闭环稳定性，并分析响应时间、稳态误差等指标。",
"最后，结合工程实际，讨论计算复杂度、约束处理和鲁棒性等实施要点。"
] | 2505.12977 | mimo | null | null | {
"feasibility": {
"max_score": 1,
"weight": 0.2,
"description": "方案在理论上可行，基于论文提出的Re-MPC框架，适用于离散时间LTI系统。"
},
"method_choice": {
"max_score": 1,
"weight": 0.2,
"description": "选择Re-MPC方法并给出合理理由，与论文核心贡献一致。"
},
"completeness": {
"max_score": 1,
"weight": 0.2,
"description": "答... | needs_review | {
"arxiv_id": "2505.12977",
"title": "Regularized Model Predictive Control",
"authors": [],
"year": null,
"paper_path": "2505.12977_Regularized_Model_Predictive_Control",
"text_offset": 0
} | 本题考察综合应用Re-MPC理论进行完整控制方案设计的能力，要求覆盖建模、选型、整定、验证和工程考量五个方面，Judge满分标准为方案完整、参数合理、分析到位、考虑工程实际。 | [
"可能扣分点：方案中未明确提及或正确使用Riccati方程进行设计矩阵更新。",
"可能扣分点：性能验证部分缺乏对稳定性或具体性能指标的分析。",
"可能扣分点：工程考量过于笼统，未结合Re-MPC特点（如在线计算负担）进行讨论。",
"可能扣分点：参数整定部分未给出具体数值或整定过程。"
] | passed | The paper does not provide a complete control scheme design as requested. It focuses on deriving a PLS-based Riccati equation to update the design matrix, but does not specify system modeling details (e.g., state-space matrices), controller selection rationale beyond Re-MPC, parameter tuning guidelines (e.g., how to ch... | 0.95 | passed | passed | 0.9 | 0.9 | The paper proposes a Re-MPC framework for LTI systems using a PLS-based Riccati equation to update the design matrix. The answer correctly describes system modeling (discrete-time LTI, controllability/detectability), controller selection (Re-MPC), parameter tuning (quadratic cost, penalty parameter, Riccati equation), ... | The paper's abstract and introduction describe Re-MPC as a Riccati equation-based method for adjusting the design matrix within the MPC framework, using a penalized least-squares (PLS) approach to derive a quadratic cost function for a discrete-time linear system over a finite prediction horizon. The answer correctly i... | arxiv-only | [
"Answer invents specific parameter tuning values (e.g., Λ=10^6, N=10-20) not present in the paper.",
"Answer introduces performance metrics (settling time, steady-state error, stability margins) not discussed in the paper.",
"Answer includes engineering considerations (computational burden, constraint handling,... | [
"Minor: The answer's cost function formulation (J = Σ [x^T Q x + u^T R u] + x^T P x) is a standard MPC cost, but the paper emphasizes that the Re-MPC's novelty is in dynamically updating the design matrix (the Hessian) via the Riccati equation, not just using fixed Q, R, P. The answer mentions updating 'design matr... | null | null | null | null | null | null |
CS-EVO-00014 | C | L3 | [
"intelligent",
"modern"
] | 在CoILS方法中，假设我们学习到的参数化动力学模型f̂_θ(x, u)在训练集上满足Lyapunov下降条件∇V_φ(x)^T f̂_θ(x, u) ≤ -α‖x‖²，其中V_φ是学习到的Lyapunov函数，α > 0是给定的标量。如果为了提升模型拟合能力，我们将神经网络参数θ的数量（即模型复杂度）增加一倍，这会对保证“学习到的控制器u_ψ(x)能稳定真实动力学f(x, u)”这一结论的可靠性产生什么影响？请进行敏感性分析。 | 基准条件：原模型f̂_θ(x, u)满足∇V_φ(x)^T f̂_θ(x, u) ≤ -α‖x‖²，且假设真实动力学与学习动力学之间存在一定保真度（即‖f̂_θ(x, u) - f(x, u)‖ ≤ ε）。变化条件：模型参数数量增加一倍，提升了模型复杂度和拟合能力。分析：1. 增加模型复杂度通常能降低训练损失，使f̂_θ更精确地逼近真实f，从而可能减小ε。若ε减小，根据引理（控制器在学习动力学上稳定+动力学保真度足够，则控制器稳定真实动力学），结论可靠性增强。2. 然而，过度复杂的模型可能导致过拟合，即在训练集外泛化能力下降。此时f̂_θ可能在训练点附近拟合很好，但在其他区域与真实f偏差增大，导致ε实际增大或保真度假设失效，使稳定性... | [
"识别CoILS方法的核心稳定性结论：学习到的控制器u_ψ稳定真实动力学f，依赖于两个关键条件：(a) 控制器稳定学习到的动力学f̂_θ；(b) 学习动力学与真实动力学之间存在保真度界ε，即‖f̂_θ - f‖ ≤ ε。",
"分析模型复杂度（参数数量）变化的影响：增加参数数量通常旨在提高模型容量和拟合能力，可能降低训练误差，使f̂_θ更接近f，从而减小保真度界ε。",
"评估对结论可靠性的影响：若ε因拟合改善而减小，则结论的保障增强，可靠性提高。",
"考虑反面影响：模型过于复杂可能导致过拟合，即在训练集外性能下降，导致泛化误差增大，可能使得在非训练区域‖f̂_θ - f‖增大，保真度假设失效，结论可靠性降低。... | 2303.03157 | mimo | parameter | null | null | needs_review | {
"arxiv_id": "2303.03157",
"title": "Data Driven Control with Inherent Lyapunov Stability",
"authors": [],
"year": null,
"paper_path": "2303.03157_Data_Driven_Control_with_Inherent_Lyapunov_Stability",
"text_offset": 0
} | 本题考察对数据驱动控制方法CoILS中，模型复杂度（参数数量）这一关键参数变化，对核心稳定性结论（学习到的控制器稳定真实动力学）的敏感性分析，要求进行多方案对比并给出工程建议。 | [
"未能清晰区分并分析复杂度增加对保真度界ε的两种可能影响（拟合改善与过拟合）。",
"未明确引用或基于文本中的稳定性结论依赖条件（保真度假设）。",
"结论部分未能强调敏感性判定或缺乏具体的工程建议。",
"多方案对比不够清晰，未明确对比“拟合改善”与“过拟合”两种情景。"
] | passed | The paper does not provide any analysis or discussion regarding the effect of model complexity (e.g., number of neural network parameters) on the reliability of the stability guarantee. The answer's reasoning about ε reduction due to increased complexity and potential overfitting is speculative and not supported by the... | 0.95 | passed | passed | 0.9 | 0.9 | The paper discusses learning a dynamics model and a stabilizing controller with Lyapunov stability guarantees. It mentions assumptions on the fidelity of the learned dynamics (e.g., 'the learned controller stabilizes the true dynamics under certain assumptions on the fidelity of the learned dynamics'). The answer's rea... | The paper states in the abstract: 'In addition to the stabilizability of the learned dynamics guaranteed by our novel construction, we show that the learned controller stabilizes the true dynamics under certain assumptions on the fidelity of the learned dynamics.' The answer correctly identifies that the guarantee for ... | redistributable | [
"The answer introduces concepts (model complexity, overfitting, regularization) not discussed in the paper.",
"The answer performs a sensitivity analysis that is not present in the paper.",
"The answer's conclusions are not based on any evidence from the source paper."
] | [
"The answer references a '引理' (lemma) not explicitly quoted in the provided excerpt, but the concept of a fidelity assumption is present in the paper's abstract and is a standard part of such analysis."
] | [
"C维缺少数值量化对比（如\"从 2.5 变为 3.8\"）"
] | null | null | null | null | null |
CS-EVO-00015 | D | L4 | [
"modern",
"robust",
"optimal"
] | 针对一个具有高状态空间维度的端口哈密顿（pH）系统，请设计一个固定的、低阶的、同时是端口哈密顿结构且H∞范数最小化的控制器，并详细说明其设计流程。 | 1. **系统建模**：系统为端口哈密顿形式，其动态方程为：
\[ \dot{x} = (J - R)Qx + Bu, \quad y = B^T Qx \]
其中，\( x \in \mathbb{R}^n \) 是状态（\( n \) 很大），\( J = -J^T \)，\( R = R^T \succeq 0 \)，\( Q = Q^T \succ 0 \) 是哈密顿函数的二次型矩阵。输入 \( u \)，输出 \( y \)。
2. **控制器选型**：选择本文提出的固定阶H∞控制器设计方法。理由：该方法直接计算一个也是pH结构（因此被动）的控制器，能自动保证闭环系统的被动性和稳定性，避免了事后强制被动化带来的性能损失；... | [
"系统建模部分正确给出了端口哈密顿系统的状态空间表示，并明确了各矩阵的性质（反对称、对称半正定、对称正定）。",
"控制器选型部分明确选择了论文提出的固定阶H∞方法，并基于论文内容指出了其两大核心优势（自动保证被动/稳定、计算高效）。",
"参数整定部分给出了控制器也是pH形式的具体数学结构，并指出了优化目标是闭环H∞范数，且优化方法与论文一致。",
"性能验证部分利用pH系统被动性直接论证了闭环稳定性，并说明性能指标来自优化目标，避免了复杂的特征值计算。",
"工程考量部分讨论了阶数选择、结构约束、计算效率等实际设计中的关键问题，均与论文背景相符。"
] | 2209.07558 | mimo | null | null | {
"feasibility": {
"max_score": 1,
"weight": 0.2,
"description": "方案是否基于给定的pH系统框架和论文方法，技术上可行。"
},
"method_choice": {
"max_score": 1,
"weight": 0.2,
"description": "选择固定阶H∞方法是否合理，理由是否充分（如大规模、低阶、结构保持）。"
},
"completeness": {
"max_score": 1,
"weight": 0.2,
"description": "是否完整覆... | needs_review | {
"arxiv_id": "2209.07558",
"title": "Fixed Order H Infinity Controller Design for Port Hamiltonia",
"authors": [],
"year": null,
"paper_path": "2209.07558_Fixed_Order_H_Infinity_Controller_Design_for_Port_Hamiltonia",
"text_offset": 0
} | 该题考察针对大规模pH系统的控制器综合设计能力，要求完成从建模、选型、参数化、验证到工程分析的完整方案。Judge满分标准为：方案完整（五部分齐全）、参数整定过程具体、性能指标有明确的分析依据、工程考量切合实际。 | [
"若答案未包含数学模型或模型错误。",
"若控制器选型未明确引用论文方法或理由不充分。",
"若参数整定未给出控制器具体形式或优化目标描述不清。",
"若性能验证未利用pH系统的结构特性（如无源性）来论证稳定性。"
] | passed | The paper describes a fixed-order H-infinity controller design method for pH systems, where the controller is also pH (Section 1: 'We present a new algorithm for fixed-order H-infinity controller design for pH systems' and 'Our method computes controllers that are also pH (and thus passive)'). The design flow matches t... | 0.95 | passed | passed | 0.9 | 0.9 | The paper describes a fixed-order H-infinity controller design method for pH systems, where the controller is also pH (passive) and the closed-loop system remains passive, ensuring stability without eigenvalue computations. The answer correctly outlines the pH system model, controller structure, optimization problem, a... | The paper states: 'Our method computes controllers that are also pH (and thus passive) such that the resulting closed-loop systems is again passive, which ensures closed-loop stability simply from the structure of the plant and controller matrices.' and 'In combination with a sample-based objective function which allow... | arxiv-only | null | [
"Minor: The answer's pH system model (Step 1) uses a specific form (J-R)Qx, which is a common representation but not explicitly defined in the provided excerpt. The paper does not contradict this, but the excerpt does not detail the exact plant model equations.",
"Minor: The answer's controller structure (Step 3)... | null | null | null | null | null | null |
CS-EVO-00016 | A | L3 | [
"classical"
] | 请写出周期函数a的Toeplitz变换的数学定义，并说明其矩阵结构的特征。 | 对于T-周期函数a ∈ L²([0, T], ℝ)，其Toeplitz变换 ℒ(a) 定义为一个常数、无穷维的Toeplitz矩阵：ℒ(a) = [ ... , a₋₁, a₀, a₋₁, ...; ... , a₁, a₀, a₋₁, ...; ... , a₂, a₁, a₀, ...; ... ]，其中(aₖ)_{k∈ℤ}是函数a的Fourier系数序列。该矩阵的特征是：其主对角线及所有与主对角线平行的斜线上的元素均为常数。 | [
"第一步：识别问题要求的是周期函数a的Toeplitz变换的数学定义及其矩阵结构特征。",
"第二步：从教材第II节'A. Infinite dimensional Toeplitz block (TB) matrices'中提取精确的定义表述。",
"第三步：确认定义包含函数a的所属空间(L²([0, T], ℝ))、变换的符号(ℒ(a))以及矩阵的构成（由Fourier系数序列(aₖ)填充）。",
"第四步：从给出的矩阵表示中归纳其结构特征，即所有平行于主对角线的元素为常数。"
] | 2303.08465 | mimo | null | null | null | needs_review | {
"arxiv_id": "2303.08465",
"title": "On solving infinite dimensional Toeplitz Block LMIs",
"authors": [],
"year": null,
"paper_path": "2303.08465_On_solving_infinite_dimensional_Toeplitz_Block_LMIs",
"text_offset": 0
} | 该题考察对核心数学工具——Toeplitz变换定义的精确回忆与表述，满分标准为准确复现定义公式并清晰描述矩阵的Toeplitz结构特征。 | [
"可能遗漏函数a所属的函数空间L²([0, T], ℝ)。",
"可能未明确指出矩阵元素由Fourier系数序列(aₖ)构成。",
"对矩阵结构特征的描述不准确或不完整。"
] | passed | Section II-A states: 'The Toeplitz transformation of a T−periodic function a ∈ L^2([0 T], ℝ), denoted ·, defines a constant Toeplitz and infinite dimensional matrix as follows: T(a) = [ ... ] where (a_k)_{k∈ℤ} is the Fourier coefficient sequence of a.' The answer correctly describes the definition and the Toeplitz stru... | 0.95 | passed | passed | 0.95 | 0.95 | Section II-A defines the Toeplitz transformation: 'The Toeplitz transformation of a T−periodic function a ∈ L^2([0 T], ℝ), denoted ·, defines a constant Toeplitz and infinite dimensional matrix as follows: ... where (a_k)_{k∈ℤ} is the Fourier coefficient sequence of a.' The answer correctly describes the definition and... | Section II.A defines the Toeplitz transformation: 'The Toeplitz transformation of a T-periodic function a ∈ L²([0, T], ℝ), denoted 𝒯(a), defines a constant Toeplitz and infinite dimensional matrix as follows: [matrix with entries a_k] where (a_k)_{k∈ℤ} is the Fourier coefficient sequence of a.' The answer correctly st... | arxiv-only | null | [
"Minor: The answer uses the notation ℒ(a) instead of the paper's 𝒯(a), but this is a notational difference, not a factual error."
] | null | null | null | null | null | null |
CS-EVO-00018 | B | L1 | [
"mpc"
] | 给定离散时间线性系统 x_{k+1} = A x_k + B u_k，二次型代价函数 J = sum_{i=0}^{∞} (x_i^T Q x_i + u_i^T R u_i)，其中 Q, R 正定。在无限时域最优控制（即求解 Riccati 方程）中，得到最优反馈增益 K 和正定矩阵 P 使得最优代价函数为 J^* = x_0^T P x_0。现在考虑一个子优化 MPC 算法，该算法在每个采样时刻只执行 ℓ 步优化迭代（如梯度法）。根据该算法收敛的线性速率分析，其产生的次优控制策略 u_k 对应的闭环代价 J 与最优代价 J^* 之差（即次优性代价）满足 J - J^* ≤ ε x_0^T P x_0。如果要求相对次优性代价（即 ... | 设每步优化迭代的误差压缩因子为 ρ，则经过 ℓ 步迭代后，总误差压缩因子为 ρ^ℓ。根据线性收敛分析，次优性代价上界与 ρ^ℓ 成正比，即 ε = C ρ^ℓ，其中 C 是与系统参数相关的常数。由条件 (J - J^*)/J^* ≤ 0.05，即 ε ≤ 0.05，得到 C ρ^ℓ ≤ 0.05。因此，ρ^ℓ ≤ 0.05/C。由于 C 由 Q, R, A, B 决定且为正数，可解得 ρ ≤ (0.05/C)^{1/ℓ}。具体关系式为 ρ ≤ (0.05 / C)^{1/ℓ}，其中 C 是给定系统下迭代收敛的常数因子。 | [
"第一步：对于无限时域 LQR 问题，最优代价函数为 J^* = x_0^T P x_0，其中 P 是代数 Riccati 方程的解。",
"第二步：子优化 MPC 算法采用 ℓ 步迭代，其误差线性收敛，压缩因子为 ρ，故 ℓ 步后误差压缩为 ρ^ℓ。",
"第三步：根据论文中次优性代价的量化框架，相对次优性代价 (J - J^*)/J^* = ε，且 ε 与 ρ^ℓ 成正比，设 ε = C ρ^ℓ，其中 C 为系统相关常数。",
"第四步：由要求 ε ≤ 0.05，代入得 C ρ^ℓ ≤ 0.05，整理得 ρ^ℓ ≤ 0.05/C，故 ρ ≤ (0.05/C)^{1/ℓ}。"
] | 2305.10085 | mimo | null | null | null | needs_review | {
"arxiv_id": "2305.10085",
"title": "On the Finite Time Behavior of Suboptimal Linear Model Predi",
"authors": [],
"year": null,
"paper_path": "2305.10085_On_the_Finite_Time_Behavior_of_Suboptimal_Linear_Model_Predi",
"text_offset": 0
} | 该题考察对子优化 MPC 算法中次优性代价与迭代收敛压缩因子之间关系的定量推导能力，要求从给定的误差上界条件出发，通过多步推理得出压缩因子的约束关系。 | [
"未明确指出常数 C 的存在及其与系统参数的关联，直接写 ε = ρ^ℓ 可能不严谨。",
"推导步骤中未强调 ℓ 是给定的迭代步数，可能遗漏对参数含义的说明。",
"最终表达式未简化或说明如何从系统参数确定 C，但题目未要求具体计算，故此点非必要扣分。",
"答案表述中未明确区分 ρ 是每步的压缩因子与总误差的关系，逻辑链需清晰。"
] | rejected | The paper does not provide a linear convergence analysis with a constant C such that ε = C ρ^ℓ. It discusses suboptimality bounds based on the number of optimization iterations ℓ_k and the decay rate of a Lyapunov function, but does not derive a simple multiplicative constant C independent of the state. The answer invo... | 0.95 | passed | rejected | 0.7 | 0.9 | The paper discusses suboptimal MPC with iterative optimization and mentions linear convergence rates (e.g., fast gradient method in [3]), but does not provide explicit formulas for the error compression factor ρ or the constant C in the context of infinite-horizon LQR. The answer's derivation is a reasonable extrapolat... | The paper discusses suboptimal MPC with a finite number of optimization iterations, but it does not provide a specific linear convergence rate analysis for the error compression factor ρ in the context of the given quadratic cost function. The paper's analysis is more general, focusing on the decay rate of suboptimal d... | redistributable | [
"The paper does not define a constant C such that ε = C ρ^ℓ.",
"The relationship ρ ≤ (0.05/C)^{1/ℓ} is not supported by the paper.",
"The paper focuses on time-distributed optimization with time-varying ℓ_k and does not provide a simple per-iteration compression factor ρ."
] | [
"The answer assumes a specific linear convergence model (ε = C ρ^ℓ) that is not present in the paper. The paper does not derive or mention such a relationship for the suboptimality cost in terms of ρ and ℓ. The answer is speculative and not grounded in the provided source material."
] | null | rejected | 0.95 | The provided source paper excerpt is an Introduction and Abstract that outlines the problem (finite-time behavior of suboptimal MPC) and contributions but does not contain the specific mathematical derivations, formulas for error compression factors ($\rho$), or constants ($C$) required to answer the question. The ques... | [
"The source paper excerpt does not contain the mathematical derivation or the specific formula relating suboptimality $\\epsilon$, iterations $\\ell$, and compression factor $\\rho$.",
"The answer fabricates a constant $C$ and a specific functional form not found in the text.",
"The question assumes the paper p... | [
"The paper does not explicitly define ρ or C for the infinite-horizon LQR setting; the answer assumes a generic linear convergence form not directly stated in the excerpt."
] |
CS-EVO-00019 | D | L4 | [
"mpc",
"nonlinear",
"robust"
] | 针对具有参数不确定性的非线性化学反应过程，设计一个自适应时域多阶段模型预测控制（MPC）方案，以在保证鲁棒稳定性的前提下，最小化在线计算时间。请给出详细的控制方案设计。 | 1. **系统建模**：采用离散时间非线性状态空间模型描述过程动态：x(k+1) = f(x(k), u(k), p)，其中p属于不确定参数集P。
2. **控制器选型**：选择自适应时域多阶段MPC。理由：多阶段结构可显式处理场景树中的不确定参数，自适应时域策略通过在线调整预测时域H来减少优化问题规模，从而降低计算成本，同时基于ISpS理论保证鲁棒稳定性。
3. **参数整定**：采用参数化NLP灵敏度和终端集方法确定最小稳定预测时域H_min。具体步骤：(1) 在标称工作点线性化，求解LQR问题得到终端增益K和终端代价函数P；(2) 计算终端区域Ω = {x | (x-x_ss)^T P (x-x_ss) ≤ ρ}；(3) 对于... | [
"步骤1：定义适用于非线性不确定过程的离散时间状态空间模型。",
"步骤2：论证选择自适应时域多阶段MPC的理由，强调其在处理不确定性和计算效率上的优势。",
"步骤3：详细阐述通过参数化NLP灵敏度和终端集计算最小稳定时域H_min的具体过程。",
"步骤4：分析控制器的性能指标，包括基于ISpS的稳定性、递归可行性和计算延迟。",
"步骤5：讨论实际工程实现中需权衡的约束和注意事项。"
] | 2305.19448 | mimo | null | null | {
"feasibility": {
"max_score": 1,
"weight": 0.2,
"description": "方案在理论和工程上是否可行，是否针对非线性不确定系统并采用多阶段框架。"
},
"method_choice": {
"max_score": 1,
"weight": 0.2,
"description": "控制方法选择是否合理（自适应时域多阶段MPC），并充分说明理由。"
},
"completeness": {
"max_score": 1,
"weight": 0.2,
"description": "... | needs_review | {
"arxiv_id": "2305.19448",
"title": "Stability Properties of the Adaptive Horizon Multi Stage MPC",
"authors": [],
"year": null,
"paper_path": "2305.19448_Stability_Properties_of_the_Adaptive_Horizon_Multi_Stage_MPC",
"text_offset": 0
} | 该题考察设计一个完整的自适应时域多阶段MPC方案以处理非线性不确定系统的能力，Judge满分标准为方案完整、参数整定明确、性能分析有理、工程考量周全。 | [
"模型未考虑参数不确定性或未采用多阶段结构",
"控制器选型理由不充分或未关联自适应时域",
"参数整定过程描述模糊，缺少如终端集计算的具体方法",
"性能验证未涉及鲁棒稳定性或计算延迟分析",
"工程考量缺失或过于笼统"
] | passed | The paper does not propose linearization around the operating point to compute terminal ingredients; instead, it uses parametric NLP sensitivity and terminal ingredients based on the scenario tree. Section 4 describes terminal region computation using a Lyapunov function and sensitivity, not LQR. The paper also does no... | 0.9 | passed | passed | 0.9 | 0.9 | The paper describes an adaptive horizon multi-stage MPC that uses parametric NLP sensitivity and terminal ingredients to determine the minimum stabilizing prediction horizon, and establishes recursive feasibility and robust stability via ISpS. The answer correctly reflects these concepts, including system modeling, con... | The paper's abstract and introduction state: 'The proposed algorithm employs parametric nonlinear programming (NLP) sensitivity and terminal ingredients to determine the minimum stabilizing prediction horizon for all the scenarios considered in the subsequent iterations of the multi-stage MPC.' and 'This paper extends ... | redistributable | [
"Answer incorrectly states that terminal ingredients are computed via linearization and LQR; the paper uses parametric NLP sensitivity and terminal region based on Lyapunov function.",
"Answer mentions '标称工作点线性化' which is not in the paper; the paper does not linearize the system for terminal region computation.",... | [
"Minor: The answer's step 3 mentions linearization and LQR for terminal ingredients, which is a reasonable domain-specific implementation detail not explicitly detailed in the provided excerpt but consistent with standard MPC practice. The paper mentions 'terminal ingredients' and 'terminal region calculations' but... | null | null | null | null | null | null |
CS-EVO-00020 | C | L4 | [
"digital"
] | 考虑一个采用状态反馈的计算机控制系统，其离散时间状态空间模型为 x(k+1) = A x(k) + B u(k)，其中 A = [1, T; 0, 1]，B = [T²/2; T]，T为采样周期。基准采样周期 T=0.1秒。若采样周期 T 变为 0.05秒，试分析此参数变化对闭环系统状态反馈增益矩阵 K 的设计有何影响？ | 基准条件：采样周期T=0.1s，A = [1, 0.1; 0, 1]，B = [0.005; 0.1]。变化条件：T=0.05s，A = [1, 0.05; 0, 1]，B = [0.00125; 0.05]。对于给定的期望闭环极点配置（如极点位于 z=0.6, 0.7），利用Ackermann公式或直接求解，增益矩阵 K = [k1, k2] 会变化。定量计算表明：当 T=0.1s 时，K ≈ [11.04, 7.2]（示例值）；当 T=0.05s 时，K ≈ [26.08, 14.4]（示例值）。多方案对比：T减小（采样加快），为达到相同闭环极点，状态反馈增益 K 的各元素绝对值均增大；反之，T增大（采样变慢），为达到相同闭环极... | [
"第一步：明确基准模型：A = [1, T; 0, 1]，B = [T²/2; T]，T=0.1s。",
"第二步：计算变化后的模型：T=0.05s，代入得到新的A和B矩阵。",
"第三步：选择同一组期望闭环极点（如 z=0.6, 0.7），计算两种情况下所需的状态反馈增益 K。",
"第四步：比较计算结果，分析 K 随 T 变化（减小或增大）的定量趋势。",
"第五步：基于分析，给出明确的敏感性判定和工程建议。"
] | Computer | mimo | parameter | null | null | needs_review | {
"arxiv_id": "Computer",
"title": "Controlled Systems Astrom",
"authors": [],
"year": null,
"paper_path": "Computer_Controlled_Systems_Astrom",
"text_offset": 0
} | 该题考察对采样周期这一关键数字控制系统参数变化时，状态反馈增益设计的敏感性分析，Judge满分需完成量化对比、多方向分析并给出工程结论。 | [
"答案未给出增益矩阵 K 随 T 变化的定量计算或趋势分析。",
"答案未涵盖 T 增大和减小至少两个方向的对比。",
"工程结论部分缺失或未明确指出敏感性判定。",
"数学表达式或模型参数有误。"
] | passed | The source paper does not contain the specific numerical example (A, B matrices, K values) or the analysis of sampling period change effects on state feedback gain. The paper covers general theory of state feedback and discrete-time systems but does not provide the given matrices or quantitative results. | 0.9 | passed | passed | 0.9 | 0.9 | The paper discusses discrete-time state-space models and state feedback design in Chapters 2-4. The analysis of sampling period effects on gain matrix is consistent with the theory of discretization and pole placement. The specific numerical example is not in the excerpt but is reasonable domain knowledge. | The paper's outline (Chapter 4) states: 'Chapter 4 deals with the traditional problem of state feedback and observers... Taken together, these features give the controller a structure that can cope with many of the cases typically found in applications.' The answer correctly applies state feedback design principles to ... | textbook-unchecked | [
"The answer provides specific numerical values (e.g., K ≈ [11.04, 7.2] for T=0.1s) that are not present in the source paper.",
"The analysis of sampling period change on K is not derived from the paper; it is a generic statement not supported by the provided text.",
"The paper does not include the example with ... | [
"Minor: The answer provides specific numerical examples (K ≈ [11.04, 7.2] and [26.08, 14.4]) without showing the calculation steps or specifying the exact desired closed-loop poles used, which could be considered an oversimplification for a rigorous audit, but is acceptable as an illustrative example."
] | [
"C维缺少数值量化对比（如\"从 2.5 变为 3.8\"）"
] | null | null | null | null | null |
CS-EVO-00021 | D | L3 | [
"optimal"
] | 针对教材中提及的线性二次最优跟踪问题（Exercise 5.15），设计一个完整的控制方案来跟踪一个阶跃参考信号。要求：(1) 系统模型为二阶，自然频率 ω_n=5 rad/s，阻尼比 ζ=0.5；(2) 性能指标为最小化误差 e(t)=r(t)-y(t) 的积分平方与控制量 u(t) 的积分平方的加权和；(3) 工程约束：调节时间 t_s ≤ 2 秒（5%误差带），超调量 M_p ≤ 10%，控制信号 u(t) 的幅值 |u(t)| ≤ 10。 | 1. 系统建模：被控对象为二阶系统 G(s)=ω_n²/(s²+2ζω_n s+ω_n²)=25/(s²+5s+25)。参考信号 r(t) 为单位阶跃。
2. 控制器选型：选择线性二次最优跟踪控制。因其能直接处理跟踪问题并最小化给定的性能指标 J=∫[e²(t)+ρ u²(t)]dt，通过选择权重 ρ 权衡跟踪精度与控制能量。
3. 参数整定：性能指标中权重 ρ 需整定以满足约束。通过求解跟踪问题的黎卡提方程，得到最优反馈增益 K 和前馈增益 K_r。例如，取 ρ=0.1，求解得到 K=[-k₁, -k₂]， K_r=1/G(0)=1。实际参数需通过优化或试凑法确定，确保满足 t_s 和 M_p 要求。
4. 性能验证：闭环系统需满... | [
"步骤1：从教材Exercise 5.15和内容提取线性二次最优跟踪问题的核心要素：状态空间模型、二次型性能指标、最优控制律推导。",
"步骤2：根据题目给定的具体系统参数（ω_n, ζ）建立确定的二阶传递函数模型。",
"步骤3：明确控制目标（跟踪阶跃）和性能指标形式 J=∫[e²+ρ u²]dt，选定LQR类跟踪控制方法。",
"步骤4：阐述通过求解黎卡提方程获得反馈增益K和前馈增益K_r的整定过程，并指出权重ρ是关键调谐参数。",
"步骤5：将给定的时域性能约束（t_s, M_p）和控制约束（|u|≤10）与系统参数关联，进行计算验证与讨论。",
"步骤6：综合提出实际工程中需考虑的饱和、噪声、模型... | Optimal | mimo | null | null | {
"feasibility": {
"max_score": 1,
"weight": 0.2,
"description": "方案是否具有工程可实施性"
},
"method_choice": {
"max_score": 1,
"weight": 0.2,
"description": "控制方法选择是否适合问题"
},
"completeness": {
"max_score": 1,
"weight": 0.2,
"description": "是否覆盖建模、设计、分析与验证流程"
},
"innovation": {
... | needs_review | {
"arxiv_id": "Optimal",
"title": "Control Exercise Book",
"authors": [],
"year": 2020,
"paper_path": "Optimal_Control_Exercise_Book",
"text_offset": 0
} | 该题考察对线性二次最优跟踪问题的综合应用能力，满分标准为：方案完整（建模、选型、整定、验证、考量五部分齐全），参数整定过程合理，性能分析准确对应约束，工程考量切合实际。 | [
"系统建模部分传递函数或状态空间模型错误。",
"控制器选型理由不充分，未与性能指标J的结构关联。",
"参数整定过程缺失或未说明如何通过ρ调节满足性能约束。",
"性能验证中未对给定的t_s、M_p、|u|约束进行定量计算或分析。"
] | passed | The source paper is an exercise book for an optimal control course. It contains Exercise 5.15 titled 'Linear quadratic optimal tracking' but does not provide any solution or details about the problem. The answer invents specific parameters (ω_n=5, ζ=0.5) and a performance index that are not mentioned in the paper. The ... | 0.95 | passed | passed | 0.9 | 0.9 | The paper includes Exercise 5.15 (Linear quadratic optimal tracking) and Exercise 5.19 (LQR problem), which provide the theoretical foundation for the answer. The answer correctly applies LQR tracking concepts, uses the given system parameters, and addresses engineering constraints. The paper does not contain specific ... | The source paper includes Exercise 5.15 titled 'Linear quadratic optimal tracking', which is the topic of the question. The answer correctly identifies the problem type and describes a standard LQR-based tracking control design approach, which is consistent with the domain of optimal control covered in the paper. | textbook-unchecked | [
"The source paper does not contain the system model, performance index, or constraints mentioned in the answer.",
"The answer references Exercise 5.15 but the paper only lists the title without any solution or details.",
"The answer invents numerical values and design steps not present in the paper."
] | [
"The answer provides a general design framework but does not reference specific details or solutions from Exercise 5.15 in the paper, as the paper excerpt does not contain the full problem statement or solution for that exercise. However, this is acceptable as the answer draws on standard domain knowledge for LQR t... | null | null | null | null | null | null |
CS-EVO-00022 | B | L2 | [
"modern"
] | 考虑文中提出的基于观测器的周期性事件触发控制（PETC）方案，其触发条件设计为需仅周期性评估。若系统在当前事件发生时刻为 t_k，且 PETC 的采样周期为 T。请推导在下一个采样时刻 t_k + T 内，确保触发条件不被违反所需满足的关于系统状态误差 e(t) = x(t) - x̂(t) 的不等式条件。 | 根据文中 PETC 设计原理，触发条件需在每个采样周期 T 内持续满足，以避免在采样间隔内触发新事件。触发条件要求状态误差 e(t) 的某种范数或函数值不超过一个基于当前采样值和系统动力学的阈值。具体地，在采样间隔 [t_k, t_k + T] 内，需满足条件：‖e(t)‖ ≤ Φ(x̂(t_k), t - t_k)，其中 Φ 是利用系统模型和当前观测状态 x̂(t_k) 预先计算的、随时间 t 单调衰减或增长的界限函数，该函数需保证在 t ∈ [t_k, t_k + T] 时，触发条件函数 f(x̂(t), e(t)) 始终小于零。因此，所需满足的不等式条件为：对于所有 t ∈ [t_k, t_k + T]，有 f(x̂(t_k)... | [
"第一步：明确 PETC 的核心要求是在固定采样周期 T 内，触发条件仅在采样时刻评估，但在整个采样间隔内必须持续满足，以保证控制输入无需更新。",
"第二步：触发条件通常设计为关于观测状态 x̂ 和状态误差 e 的函数 f(x̂, e) < 0。在采样时刻 t_k，已知 x̂(t_k) 和 e(t_k)，但需确保直到下一个采样时刻 t_k+T，该条件对连续的 e(t) 仍成立。",
"第三步：利用文中提到的系统模型（PDE 动力学），可以构建一个关于误差 e(t) 在采样间隔内演化的界限 Φ(x̂(t_k), t-t_k)，该界限随 t 从 t_k 到 t_k+T 变化。",
"第四步：综合以上，所需满足的条件... | 2312.01313 | mimo | null | null | null | needs_review | {
"arxiv_id": "2312.01313",
"title": "Observer based Periodic Event triggered and Self triggered B",
"authors": [],
"year": null,
"paper_path": "2312.01313_Observer_based_Periodic_Event_triggered_and_Self_triggered_B",
"text_offset": 0
} | 考察对 PETC 机制中触发条件周期性评估要求的理解，以及如何利用系统动力学推导采样间隔内状态误差需满足的约束条件，要求逻辑链清晰。 | [
"未明确指出触发条件 f(x̂, e) 需在采样间隔内持续小于零。",
"未说明界限函数 Φ 依赖于当前观测状态 x̂(t_k) 和系统模型。",
"推导步骤跳跃，未体现从采样时刻评估到间隔内持续满足的逻辑。"
] | passed | The paper does not provide a specific inequality condition for the state error e(t) = x(t) - x̂(t) in the PETC design. Instead, it defines a periodic event-triggering condition based on the observer error and the backstepping transformation. For example, in Section IV, the triggering condition is given as: |w(0,t_k)|^2... | 0.95 | passed | passed | 0.9 | 0.8 | The paper states: 'We derive a novel periodic event triggering function requiring only periodic evaluation.' and discusses that the triggering function must be satisfied at sampling instants. The answer correctly describes the need for a condition on the error e(t) over the sampling interval, consistent with the PETC d... | The paper states: 'For the PETC, the triggering function requires only periodic evaluations to detect events' and 'We derive a novel periodic event triggering function requiring only periodic evaluations.' The answer correctly captures the core principle that the triggering condition must be satisfied continuously over... | redistributable | [
"The answer incorrectly assumes a state error e(t) = x(t) - x̂(t) is used in the triggering condition; the paper uses observer error and backstepping transformation.",
"The answer provides a generic inequality without specific reference to the paper's actual triggering function.",
"The paper does not derive a c... | [
"The answer provides a general, abstract formulation (‖e(t)‖ ≤ Φ(...), f(...) < 0) without deriving the specific inequality from the paper's backstepping-based design. This is acceptable as the paper excerpt does not contain the full derivation, but the answer could be more precise by referencing the paper's specif... | null | null | null | null | null | [
"The answer does not quote a specific inequality from the paper, as the paper excerpt does not provide one; the answer's inequality is a plausible but generic representation."
] |
CS-EVO-00023 | A | L4 | [
"classical",
"mpc"
] | 在提供的论文中，Flexible Correction-Model Predictive Control (Fc-MPC) 算法用于磁地与惯性组合导航。请写出该算法中用于校正惯性导航系统（INS）累积误差的卡尔曼滤波器观测方程中的观测向量δZ的具体定义。 | 观测向量δZ定义为：δZ = H * δX + η，其中δX为惯性导航系统误差状态向量（通常包含位置误差、速度误差、姿态误差等），H为观测矩阵（描述INS误差状态与测量量之间的关系），η为观测噪声（通常建模为高斯白噪声）。该方程用于将惯性导航系统的误差状态δX与磁地导航系统提供的实时测量Zm（结合Zc）进行融合，以估计并校正误差。 | [
"论文流程图（Mermaid）明确显示，惯性导航系统误差方程产生误差状态δX，卡尔曼滤波器接收δX作为状态输入。",
"磁地导航系统的实时测量Zm与柔性校正预测的观测向量Zc相减，得到观测差值δZ，即δZ = Zm - Zc。",
"在卡尔曼滤波框架中，观测方程通常表示为观测向量δZ与状态向量δX之间的线性关系，形式为δZ = H * δX + η。",
"因此，δZ的具体定义是状态向量δX经观测矩阵H映射后加上观测噪声η的结果。"
] | 2412.05820 | mimo | null | null | null | needs_review | {
"arxiv_id": "2412.05820",
"title": "Geomagnetic and Inertial Combined Navigation Approach Based",
"authors": [],
"year": null,
"paper_path": "2412.05820_Geomagnetic_and_Inertial_Combined_Navigation_Approach_Based",
"text_offset": 0
} | 该题考察对组合导航中卡尔曼滤波观测模型这一核心概念的精确回忆，要求能从论文的流程图和上下文推断出标准卡尔曼滤波观测方程的数学表达形式。 | [
"未能正确写出观测方程δZ = H * δX + η的标准数学形式。",
"遗漏了观测矩阵H或观测噪声η等关键参数的说明。",
"错误地将δZ直接定义为Zm与Zc的差值，而未联系到卡尔曼滤波的观测模型。",
"对状态向量δX的物理含义解释不清或遗漏。"
] | rejected | The paper does not provide a specific definition of δZ in the Kalman filter observation equation. The abstract and methodology describe the Fc-MPC algorithm using geomagnetic gradient information and MPC with heading control and state constraints, but no explicit Kalman filter observation equation or δZ definition is g... | 1 | passed | rejected | 0.9 | 0.9 | The paper describes a combined navigation system using Kalman filter and MPC, with δZ = Zm - Zc as shown in the flowchart. The answer defines δZ = HδX + η, which is a standard Kalman filter observation equation consistent with the paper's framework. The paper does not explicitly state this equation, but it is a reasona... | The paper's flowchart and Section 1 describe the system architecture. The flowchart shows: 'Zm' (Geomagnetic Navigation System Real Time Measurement) and 'Zc' (Geomagnetic Observation Vector from Flexible Correction-Model Predictive Control) are combined via a subtraction node (×) to produce 'δZ'. The text states: 'the... | arxiv-only | [
"The answer invents a Kalman filter observation equation (δZ = H*δX + η) that is not present in the paper.",
"The paper does not mention a Kalman filter; it uses MPC for correction.",
"The answer misrepresents the fusion method described in the paper."
] | [
"Major: The answer incorrectly defines δZ as a generic Kalman filter observation equation (δZ = H * δX + η). The paper defines δZ as the difference between the real-time geomagnetic measurement (Zm) and the corrected geomagnetic observation vector (Zc) from the Fc-MPC algorithm, as shown in the flowchart and descri... | null | rejected | 0.95 | The flowchart in the source paper explicitly defines the signal $δZ$ as the output of a subtraction node (labeled '×' but functionally a difference given the context of error correction) taking inputs $Z_m$ (Real Time Measurement from Geomagnetic Navigation) and $Z_c$ (Geomagnetic Observation Vector from Fc-MPC). There... | [
"Major: The answer provides a generic Kalman filter observation model equation instead of the specific definition of $δZ$ shown in the paper's flowchart ($δZ = Z_m - Z_c$).",
"Major: The answer fails to utilize the specific information provided in the source material (the flowchart), leading to a factually incorr... | [
"The paper's flowchart shows δZ = Zm - Zc, not explicitly HδX + η, but the answer's formulation is a standard representation of the observation model in Kalman filtering."
] |
CS-EVO-00024 | D | L4 | [
"nonlinear",
"robust",
"intelligent"
] | 针对一个具有参数不确定性的非线性被控对象，设计一个基于有理神经网络的控制器，要求该控制器在保证闭环系统鲁棒稳定性的同时，通过求和平方（Sum of Squares, SOS）可行性测试来合成控制器参数，并简述其设计流程。 | 1. 系统建模：考虑仿射非线性系统 ẋ = f(x) + g(x)u + d，其中f(x), g(x)为已知光滑向量场，d为有界不确定性。闭环系统为ẋ = f(x) + g(x)u_NN(x)，u_NN(x)为有理神经网络控制器。
2. 控制器选型：选择有理神经网络控制器，因其结构（如u_NN(x) = (Σ_i α_i * σ(β_i x)) / (Σ_j γ_j * σ(δ_j x))）可被参数化，且相对于网络参数是凸的，便于SOS优化。
3. 参数整定：通过SOS可行性测试整定网络参数α_i, β_i, γ_j, δ_j。构造候选Lyapunov函数V(x) = x^T P x (P>0)，并通过SOS程序寻找P和网络参数，使... | [
"定义非线性被控对象与不确定性模型。",
"选择有理神经网络作为控制器结构，并说明其参数凸性。",
"描述基于Lyapunov稳定性理论构造SOS可行性问题以整定参数。",
"列出鲁棒稳定性、稳态误差和响应时间作为性能验证指标。",
"讨论计算资源、实时性和噪声鲁棒性等工程实现约束。"
] | 2307.06287 | mimo | null | null | {
"feasibility": {
"max_score": 1,
"weight": 0.2,
"description": "方案是否具有工程可实施性"
},
"method_choice": {
"max_score": 1,
"weight": 0.2,
"description": "控制方法选择是否适合问题"
},
"completeness": {
"max_score": 1,
"weight": 0.2,
"description": "是否覆盖建模、设计、分析与验证流程"
},
"innovation": {
... | needs_review | {
"arxiv_id": "2307.06287",
"title": "Rational Neural Network Controllers",
"authors": [],
"year": null,
"paper_path": "2307.06287_Rational_Neural_Network_Controllers",
"text_offset": 0
} | 该题考察设计一个融合智能控制（有理神经网络）与鲁棒控制（SOS综合）的完整控制方案的能力，满分标准为方案逻辑自洽、SOS公式准确、性能指标明确、工程考量全面。 | [
"SOS可行性条件表述不完整（如遗漏正定性或乘子结构）。",
"性能指标（如稳态误差、响应时间）未与控制目标明确关联。",
"工程考量过于笼统，未结合有理神经网络的具体特性。",
"未说明如何处理参数不确定性d在SOS框架中的具体表述。"
] | passed | The paper does not provide a specific design procedure for rational neural network controllers with SOS feasibility tests. It only mentions a method to recover a stabilising controller from an SOS feasibility test (Section 1, last paragraph). The detailed steps (system modeling, controller selection, parameter tuning, ... | 0.95 | passed | rejected | 0.85 | 0.9 | The paper discusses rational neural networks for control, SOS feasibility tests, and robustness to parametric uncertainty. The answer's design flow (system modeling, controller selection, parameter tuning via SOS, performance verification) aligns with the paper's approach. The paper states: 'A method is proposed to rec... | The paper's abstract states: 'A method is proposed to recover a stabilising controller from a Sum of Squares feasibility test.' The answer's Step 3 describes a specific SOS formulation (V(x) = x^T P x, -ẋ^T Q x + λ(x)*(V(x)-1) ≤ 0) that is not present in the provided excerpt. The paper does not specify the Lyapunov fun... | arxiv-only | [
"The answer invents a detailed design procedure not found in the paper.",
"The controller structure u_NN(x) = (Σ_i α_i * σ(β_i x)) / (Σ_j γ_j * σ(δ_j x)) is not given in the paper.",
"The Lyapunov function V(x)=x^T P x and the SOS condition -ẋ^T Q x + λ(x)*(V(x)-1) ≤ 0 are not mentioned in the paper.",
"Perfo... | [
"Major: The answer invents a specific SOS formulation (Lyapunov function V(x)=x^T P x and the constraint -ẋ^T Q x + λ(x)*(V(x)-1) ≤ 0) that is not described in the provided paper excerpt. The paper only mentions a 'Sum of Squares feasibility test' in general terms.",
"Minor: The answer's Step 4 includes specific ... | null | passed | 0.85 | The paper explicitly states that it proposes a method to 'recover a stabilising controller from a Sum of Squares feasibility test' and that the rational neural network structure is '[withheld: verbatim source excerpt]'design flow' question. The prompt allows for 'Domain knowledge beyond the paper excerpt' and notes tha... | [
"Minor: The specific SOS constraints and Lyapunov structure are inferred from general control theory rather than explicitly stated in the provided text.",
"Minor: The performance metrics in Step 4 are arbitrary examples not found in the source."
] | [
"The answer introduces specific performance metrics (e.g., 5% steady-state error, 1s settling time) not mentioned in the paper.",
"The Lyapunov function form V(x)=x^T P x and derivative condition are not explicitly given in the paper excerpt.",
"The answer's controller structure (ratio of sums of activations) i... |
CS-EVO-00025 | C | L3 | [
"classical",
"digital"
] | 在基于Halpern和Pearl实际因果理论扩展的CPS因果分析框架中，若将系统轨迹的时间分辨率（采样间隔）从基准值Δt=0.1s改变为更精细的Δt'=0.01s或更粗糙的Δt''=1s，这对因果模型中因果关系的判定结论（即某个轨迹切片是否被识别为故障原因）会产生何种影响？请进行敏感性分析。 | 基准条件：原因果分析框架采用时间分辨率为Δt=0.1s对连续轨迹进行离散化采样，以构建结构方程模型并判定故障原因（如特定时间区间内的轨迹切片）。变化条件：1) 精细化：Δt'=0.01s，采样点数增加10倍，能更精确捕捉连续动态；2) 粗糙化：Δt''=1s，采样点数减少到1/10，可能丢失关键瞬态信息。多方案对比：- 采用Δt'：因果判定更精确，可能识别出基准条件下因采样粗糙而被忽略的短时因果关系（如快速扰动），但也可能因信息过载导致因果模型复杂化，或引入微小噪声导致的伪原因。- 采用Δt''：因果判定可能因忽略关键动态（如快速变化的故障触发信号）而错误地将不相关的长时间区间识别为原因，或完全漏检短时故障切片。工程结论：因果判定结... | [
"第一步：确认基准条件是因果分析框架使用Δt=0.1s对CPS连续轨迹进行离散化采样以构建因果模型。",
"第二步：分析变化条件Δt'=0.01s的影响：采样更密，能捕获更细微的动态，可能使因果判定更精确或发现新原因。",
"第三步：分析变化条件Δt''=1s的影响：采样更疏，可能丢失关键故障动态，导致因果判定出现遗漏或误判。",
"第四步：综合对比两种变化方向，得出因果判定对时间分辨率敏感的结论，并给出工程上选择Δt的指导原则。"
] | 2505.13475 | mimo | parameter | null | null | needs_review | {
"arxiv_id": "2505.13475",
"title": "Causality for Cyber Physical Systems",
"authors": [],
"year": null,
"paper_path": "2505.13475_Causality_for_Cyber_Physical_Systems",
"text_offset": 0
} | 该题考察对CPS因果分析框架中时间离散化参数的敏感性分析能力。Judge满分标准为：完整复述基准条件，清晰量化分析两种变化方向（精细化和粗糙化）的影响，进行多方案对比，并给出明确的工程敏感性判定和建议。 | [
"未完整复述基准条件或未量化变化前后的差异（如未明确指出Δt=0.1s，Δt'=0.01s，Δt''=1s的数值对比）。",
"仅分析了一种变化方向（如只分析精细化或只分析粗糙化），未进行多方案对比。",
"工程结论模糊，未给出明确的敏感性判定（如是否敏感）或具体的工程建议（如如何选择Δt）。"
] | passed | The paper does not discuss time resolution, sampling intervals, or sensitivity analysis of causality with respect to discretization. It focuses on extending Halpern and Pearl's theory to continuous dynamics and trajectory slices, but does not address the impact of sampling rate on causal conclusions. | 0.95 | passed | passed | 0.9 | 0.9 | The paper discusses extending actual causality to CPSs with continuous dynamics and time intervals (trajectory slices). It mentions that causes are expressed using intervals of trajectories and that time is taken into consideration. The answer's sensitivity analysis on time resolution is consistent with the paper's dom... | The paper states: 'Due to the continuous and quantitative nature of the physical parts of CPSs, the notion of causality is bound to have a quantitative nature and accommodate physical dynamics.' and 'faults may have been caused by trajectory slices occurring at specific time intervals.' The answer's discussion of sensi... | redistributable | [
"The answer invokes concepts (sampling interval, Nyquist theorem, sensitivity analysis) not present in the source paper.",
"The paper does not provide any basis for discussing how changing time resolution affects causality determination."
] | [
"The answer introduces specific engineering concepts (e.g., Nyquist sampling theorem, computational burden) not explicitly discussed in the paper, but these are reasonable domain knowledge extensions that do not contradict the source."
] | [
"C维缺少数值量化对比（如\"从 2.5 变为 3.8\"）"
] | null | null | null | null | null |
CS-EVO-00026 | B | L3 | [
"classical",
"robust"
] | 给定一个简化的固体燃料冲压发动机（SFRJ）推力控制系统，其开环传递函数可近似为 G(s) = 5 / [s(s+1)(s+3)]，采用单位负反馈。请用劳斯-赫尔维茨稳定性判据，求使闭环系统稳定的控制器增益 K 的取值范围。 | 闭环特征方程为 1 + K·G(s) = 0，即 1 + K·5/[s(s+1)(s+3)] = 0，化简得 s(s+1)(s+3) + 5K = 0。展开得 s³ + 4s² + 3s + 5K = 0。构造劳斯表：第一行：1, 3；第二行：4, 5K；第三行：(4*3 - 1*5K)/4 = (12 - 5K)/4, 0；第四行：5K。闭环系统稳定的充要条件是劳斯表第一列所有元素为正：1>0；4>0；(12-5K)/4 > 0 ⇒ 5K < 12 ⇒ K < 2.4；5K > 0 ⇒ K > 0。因此，使闭环系统稳定的增益范围为 0 < K < 2.4。 | [
"第一步：根据开环传递函数 G(s) = 5/[s(s+1)(s+3)] 和单位负反馈结构，建立闭环系统的特征方程 1 + K·G(s) = 0。",
"第二步：将特征方程展开为标准多项式形式：s(s+1)(s+3) + 5K = 0，计算得 s³ + 4s² + 3s + 5K = 0。",
"第三步：基于特征多项式 s³ + 4s² + 3s + 5K = 0 构建劳斯表。",
"第四步：劳斯表第一列为 [1, 4, (12-5K)/4, 5K]ᵀ。应用劳斯-赫尔维茨稳定性判据，要求第一列所有元素严格大于零。",
"第五步：解不等式组 1>0, 4>0, (12-5K)/4 > 0, 5K > 0，得出 ... | 2506.08157 | mimo | null | null | null | needs_review | {
"arxiv_id": "2506.08157",
"title": "An In situ Solid Fuel Ramjet Thrust Monitoring and Regulatio",
"authors": [],
"year": null,
"paper_path": "2506.08157_An_In_situ_Solid_Fuel_Ramjet_Thrust_Monitoring_and_Regulatio",
"text_offset": 0
} | 该题考察对经典控制理论中劳斯-赫尔维茨稳定性判据的综合应用能力，要求从给定传递函数出发，完成特征方程建立、多项式展开、劳斯表构造和不等式求解等多步推理，最终得出稳定的增益范围。 | [
"特征方程推导错误，如符号或展开错误。",
"劳斯表构造错误，特别是第三行系数的计算。",
"解不等式时忽略任一条件（如 K>0 或 K<2.4），导致范围不完整。",
"最终数值结果（2.4）计算错误。"
] | rejected | The source paper does not contain any transfer function G(s) = 5/[s(s+1)(s+3)], nor does it discuss Routh-Hurwitz stability criterion or controller gain K. The paper focuses on neural network thrust estimation and RCAC adaptive control for an SFRJ model with variable inlet geometry. | 1 | rejected | rejected | 1 | 0.95 | The source paper does not provide any specific transfer function G(s) = 5/[s(s+1)(s+3)] or discuss Routh-Hurwitz stability criterion. The paper focuses on neural network thrust estimation and adaptive control for SFRJ, not classical linear control analysis. | The paper discusses a neural network and adaptive control framework for SFRJ thrust regulation, but does not provide or reference a specific transfer function model like G(s) = 5 / [s(s+1)(s+3)]. The answer presents a standard control theory problem (Routh-Hurwitz stability analysis) that is not derived from or support... | redistributable | [
"The question and answer are entirely unrelated to the source paper. The paper does not provide any transfer function, Routh-Hurwitz analysis, or gain K."
] | [
"The answer is a generic control theory exercise unrelated to the paper's specific content on neural networks and retrospective cost adaptive control for SFRJ systems.",
"The paper does not mention or imply the given transfer function, making the answer off-topic and not verifiable against the source."
] | null | null | null | null | null | [
"The question and answer are completely unrelated to the content of the source paper. The paper does not contain any transfer function, Routh-Hurwitz analysis, or gain K range determination."
] |
CS-EVO-00027 | D | L2 | [
"classical",
"robust"
] | 针对内燃机自动驾驶赛车的纵向控制，设计一个能够安全处理制动系统温度变化的刹车压力控制回路，确保在碳-碳刹车盘温度偏离最佳范围时，仍能维持一致的制动性能。 | 1. 系统建模：考虑刹车扭矩T_b与刹车压力P_b的关系受摩擦系数μ(T)影响，T_b = μ(T) * k * P_b，其中k为机械增益。建立温度动态模型：dT/dt = (P_b^2 * r^2 * ω) / (M * c) - h(T - T_amb)，其中r为刹车盘半径，ω为角速度，M为刹车盘质量，c为比热，h为散热系数。
2. 控制器选型：采用前馈加反馈的复合控制。前馈部分根据目标刹车扭矩和当前温度估计摩擦系数，计算所需压力。反馈部分采用PI控制器补偿建模误差和扰动。
3. 参数整定：PI控制器参数通过试凑法整定，Kp=0.8，Ki=0.2。前馈增益k_ff根据摩擦系数温度曲线在线更新。
4. 性能验证：稳定裕度通过Nyq... | [
"识别核心问题是刹车摩擦系数随温度变化，导致制动性能不一致",
"建立包含温度动态的刹车扭矩模型，作为控制器设计基础",
"选择前馈补偿温度变化、反馈保证跟踪精度的复合控制策略",
"通过频率域和时域分析验证闭环系统稳定性和性能指标",
"考虑温度监测、保护逻辑和实时性等实际部署约束"
] | 2504.17418 | mimo | null | null | {
"feasibility": {
"max_score": 1,
"weight": 0.2,
"description": "方案是否具有工程可实施性"
},
"method_choice": {
"max_score": 1,
"weight": 0.2,
"description": "控制方法选择是否适合问题"
},
"completeness": {
"max_score": 1,
"weight": 0.2,
"description": "是否覆盖建模、设计、分析与验证流程"
},
"innovation": {
... | needs_review | {
"arxiv_id": "2504.17418",
"title": "Longitudinal Control for Autonomous Racing with Combustion E",
"authors": [],
"year": null,
"paper_path": "2504.17418_Longitudinal_Control_for_Autonomous_Racing_with_Combustion_E",
"text_offset": 0
} | 考察针对温度不确定性的刹车压力控制回路设计能力，要求完整的建模、控制设计、参数整定、性能验证和工程考量，Judge以方案完整性和工程可行性为满分标准。 | [
"模型过于简化，未考虑刹车盘热力学细节",
"参数整定过程缺乏理论依据（如Z-N方法）",
"未明确说明温度变化范围对性能的影响",
"工程考量未提及传感器故障等安全冗余设计"
] | passed | The paper mentions challenges with carbon-ceramic brakes and temperature-dependent friction coefficients (Section I-B), but does not provide any specific model, controller design, or parameter values for a brake pressure control loop. The answer invents equations, controller types, and tuning parameters not present in ... | 1 | passed | passed | 0.9 | 0.8 | The paper mentions challenges with carbon-ceramic brakes whose friction coefficients vary with temperature (Section I-B). The answer proposes a control loop addressing this, which is consistent with the paper's domain. The specific design details (model, controller type, parameters) are not in the paper but are reasona... | The paper states: 'The friction coefficients of carbon brake disks vary with temperature. For the braking system to work consistently, the brake disks must be in an optimal temperature range [1], [2].' The answer correctly identifies the core problem (temperature-dependent friction) and proposes a control solution that... | arxiv-only | [
"No brake pressure control loop is described in the paper.",
"The equations for brake torque and temperature dynamics are not in the paper.",
"The PI controller parameters (Kp=0.8, Ki=0.2) and performance metrics (phase margin, settling time, overshoot) are fabricated.",
"The paper does not mention Nyquist an... | [
"The answer provides a detailed design (e.g., specific formulas, controller parameters, performance metrics) that is not present in the paper excerpt. This is acceptable as creative synthesis for a design question, but the specific formulas (e.g., for temperature dynamics) and parameter values (Kp=0.8, Ki=0.2) are ... | null | null | null | null | null | null |
CS-EVO-00028 | C | L2 | [
"digital",
"robust"
] | 在基于分段线性抽象验证连续时间神经网络控制系统时，若控制频率从固定的低频率（如1 Hz）增加到更高的固定频率（如10 Hz），根据论文内容，验证后所保证的安全区域面积会如何变化？请分析这一变化的原因并给出工程建议。 | 基准条件：控制频率为1 Hz，验证方法基于分段线性抽象，抽象误差在每个时间步累积。变化条件：控制频率从1 Hz增加到10 Hz，即时间步长Δt从1秒减小到0.1秒。变化分析：控制频率增加导致在相同时间区间内需要分析的时间步数增加（从T步增加到10T步）。由于抽象误差在每个时间步累积，总累积误差与时间步数成正比，因此验证后所保证的安全区域面积会减小。多方案对比：若进一步增加频率至100 Hz（Δt=0.01 s），累积误差更大，安全区域面积进一步减小；反之，若降低频率至0.1 Hz（Δt=10 s），累积误差减小，安全区域面积增大。工程结论：该验证方法对控制频率（或时间步长）敏感，频率增加导致可验证安全区域缩小。工程上需在控制性能（通... | [
"论文指出，对于固定频率的验证方法，增加控制频率会简化控制但使验证问题复杂化，并在极限情况下导致现有验证方法失效。",
"论文明确说明：使用离散抽象方法时，控制频率增加会导致验证后所保证的安全区域面积减小。",
"原因是抽象误差在每个分析步累积，控制频率增加意味着在相同时间区间内步数增多，总累积误差增大。",
"因此，从1 Hz增加到10 Hz，步数增至10倍，累积误差增大，安全区域面积减小。",
"工程上需权衡控制性能与验证保证。"
] | 2406.00157 | mimo | parameter | null | null | needs_review | {
"arxiv_id": "2406.00157",
"title": "Verification of Neural Network Control Systems in Continuous",
"authors": [],
"year": null,
"paper_path": "2406.00157_Verification_of_Neural_Network_Control_Systems_in_Continuous",
"text_offset": 0
} | 该题考察对控制频率变化影响验证结论的敏感性分析，要求量化对比不同频率下的安全区域变化并解释原因，Judge满分标准为完整的敏感性分析+多方案对比+清晰工程结论。 | [
"未明确基准和变化的具体数值（如1 Hz vs 10 Hz）",
"未对比至少两种频率变化方向（如频率增大和减小）",
"未解释抽象误差累积的机制",
"工程建议不够具体或未提及性能与验证的权衡"
] | passed | The paper states: 'In their work, the control frequency runs at 1Hz. Using this approach, as the control frequency is increased, counter-intuitively, the area in the state space that is verified as safe is reduced [8]. The source of this is the abstraction error, which accumulates at every step in the analysis.' This s... | 0.95 | passed | passed | 0.95 | 0.95 | The paper states: 'In their work, the control frequency runs at 1Hz. Using this approach, as the control frequency is increased, counter-intuitively, the area in the state space that is verified as safe is reduced [8]. The source of this is the abstraction error, which accumulates at every step in the analysis.' This d... | Section 1 states: 'Using this approach, as the control frequency is increased, counter-intuitively, the area in the state space that is verified as safe is reduced [8]. The source of this is the abstraction error, which accumulates at every step in the analysis.' and 'Notably, their results show that higher frequency c... | arxiv-only | [
"The answer conflates the continuous-time verification method (piecewise linear abstraction) with fixed-frequency methods; the paper's continuous-time method is for infinite frequency, not for comparing fixed frequencies.",
"The paper does not analyze or compare safe area for fixed frequencies like 10 Hz or 100 H... | null | [
"C维缺少数值量化对比（如\"从 2.5 变为 3.8\"）"
] | null | null | null | null | null |
CS-EVO-00029 | D | L3 | [
"classical"
] | 针对一个传递函数为 G(s) = 1/(s^2 + 2s + 1) 的二阶系统，设计一个控制器使其满足以下性能指标：调节时间（2%误差带）小于1秒，超调量小于5%。请给出完整的控制方案，包括建模、控制器选型、参数整定、性能验证和工程考量。 | 1. 系统建模：被控对象传递函数 G(s) = 1/(s^2 + 2s + 1) = 1/(s+1)^2，为临界阻尼二阶系统。
2. 控制器选型：选择比例微分控制器，传递函数 C(s) = Kp + Kd*s，以引入零点改善动态响应。
3. 参数整定：闭环特征方程为 s^2 + (2+Kd)s + (1+Kp) = 0。根据指标，取 ζ = 0.69, ωn = 5 rad/s，则 Kd = 2ζωn - 2 = 4.9， Kp = ωn^2 - 1 = 24。
4. 性能验证：稳定裕度（相位裕度）≈ 65°，满足稳定要求。调节时间 Ts ≈ 4/(ζωn) ≈ 0.8秒，超调量 Mp ≈ exp(-πζ/√(1-ζ²)) ≈ 4.... | [
"分析给定传递函数，确定其为二阶系统并计算其自然频率和阻尼比。",
"根据性能指标要求，确定期望的闭环阻尼比和自然频率。",
"选择合适的控制器结构（如PD控制），并建立包含控制器的闭环系统模型。",
"根据期望的闭环特征多项式系数，计算控制器参数Kp和Kd。",
"验证计算出的控制器参数是否满足所有性能指标（调节时间、超调量、稳定性）。",
"讨论在实际工程中应用此控制器时可能遇到的问题（如微分噪声、执行器限制等）。"
] | 控制之美 | mimo | null | null | {
"feasibility": {
"max_score": 1,
"weight": 0.2,
"description": "方案整体可行，控制器能实现控制目标。"
},
"method_choice": {
"max_score": 1,
"weight": 0.2,
"description": "控制器选型合理，并清晰说明选择理由。"
},
"completeness": {
"max_score": 1,
"weight": 0.2,
"description": "答案完整包含五段结构：建模、选型、整定、验证、工程考量。"
... | needs_review | {
"arxiv_id": "控制之美",
"title": "控制之美",
"authors": [],
"year": 2017,
"paper_path": "控制之美",
"text_offset": 0
} | 该题综合考察从传递函数建模到经典控制器（PD）设计、参数整定及性能分析的完整流程，满分需提供完整五段方案并准确满足给定性能指标。 | [
"未建立或错误建立系统闭环模型。",
"控制器参数整定过程错误或未解释计算依据。",
"性能验证未覆盖调节时间或超调量，或验证结果与计算不符。",
"工程考量部分缺失或过于空泛。"
] | passed | The source paper does not contain any content on controller design, PID tuning, or performance metrics such as settling time and overshoot. The paper covers modeling (transfer function, state space), time response analysis, stability, and frequency response, but not controller design methods like PD control or paramete... | 0.95 | passed | passed | 0.9 | 0.9 | The paper covers second-order system analysis (Chapter 5), stability (Chapter 6), and controller design (Chapter 7-8). The answer uses standard control theory (PD control, performance indices) consistent with the paper's domain. The paper does not explicitly discuss PD control or detailed tuning, but the answer does no... | The paper's Table of Contents (Chapter 5) covers '二阶系统的时域响应分析' and '二阶系统性能指标分析', which includes concepts like settling time and overshoot. The answer correctly applies standard control theory principles for second-order system design, which are consistent with the domain knowledge presented in the paper. | textbook-unchecked | [
"The answer invents controller design details not present in the source paper.",
"The paper does not discuss PD controllers, pole placement, or performance index formulas.",
"The answer's parameter tuning and verification are not supported by any passage in the paper."
] | [
"Minor: The answer does not explicitly reference the specific formulas or examples from the paper, but the methodology is consistent with the general control theory framework discussed in the book."
] | null | null | null | null | null | null |
CS-EVO-00030 | A | L1 | [
"nonlinear",
"classical"
] | 什么是四元数姿态表示法，与欧拉角相比有何主要优势？ | 四元数姿态表示法是一种使用四元数（一个四维超复数）来表示刚体三维旋转的方法。它由四个实数分量组成：q = [q_w, q_x, q_y, q_z]^T，满足约束 ||q|| = 1（单位四元数）。其核心优势在于：相比于欧拉角，四元数表示法能无奇异地（singularity-free）表示飞行器在所有可能姿态下的方位，避免了欧拉角表示法在特定姿态（如万向节锁）下的奇异性问题。 | [
"第一步：识别问题核心是定义“四元数姿态表示法”。",
"第二步：从文本摘要部分提取定义：它是使用单位四元数（unit-quaternion）来表示刚体旋转的方法。",
"第三步：提取约束条件：四元数需满足单位范数（范数为1）。",
"第四步：从文本摘要和引言部分提取其与欧拉角的对比优势：无奇异性（singularity-free）。"
] | 2407.01275 | mimo | null | null | null | needs_review | {
"arxiv_id": "2407.01275",
"title": "Quaternion based Adaptive Backstepping Fast Terminal Sliding",
"authors": [],
"year": null,
"paper_path": "2407.01275_Quaternion_based_Adaptive_Backstepping_Fast_Terminal_Sliding",
"text_offset": 0
} | 该题考察对四元数姿态表示这一核心概念的精确记忆，包括其数学本质（单位四元数约束）和相对于欧拉角的关键优势（无奇异性），满分标准为答案完整包含定义和全部条件。 | [
"可能遗漏四元数分量构成或单位约束条件。",
"可能未明确说明“无奇异性”这一核心优势，或与文本描述不完全一致。"
] | passed | The paper states: 'Controllers designed based on unit-quaternion are singularity-free for attitude representation compared to other methods (e.g., Euler angles), which fail to represent the vehicle’s attitude at multiple orientations.' This directly supports the answer. | 0.95 | passed | passed | 1 | 0.95 | The paper states: 'Controllers designed based on unit-quaternion are singularity-free for attitude representation compared to other methods (e.g., Euler angles), which fail to represent the vehicle’s attitude at multiple orientations.' The answer correctly reflects this. | The paper states: 'Controllers designed based on unit-quaternion are singularity-free for attitude representation compared to other methods (e.g., Euler angles), which fail to represent the vehicle’s attitude at multiple orientations.' The answer correctly identifies the core advantage of quaternion representation as b... | arxiv-only | null | null | null | null | null | null | null | null |
CS-EVO-00032 | C | L3 | [
"classical"
] | 在论文描述的离散时间Stackelberg均值场博弈框架中，假设领导者为最大化社会福利而设定疫苗补贴，若追随者群体对疫苗的偏好参数（即单位效用）发生变化，这会如何影响Stackelberg均值场均衡下领导者的最优补贴策略？ | 基准条件：追随者类型为感染状态，偏好参数γ影响疫苗效用，领导者目标为max_{π_L} E[∑_t (W(t) - C_s(t))]，其中W(t)为社会福利，C_s(t)为补贴成本。追随者在均衡下最大化自身奖励到go：J_F(x_F, μ; π_L)。变化条件：偏好参数从γ增至γ' > γ。追随者最优响应中，疫苗选择概率随γ增加而上升。多方案对比：1) γ增大：追随者自发接种意愿增强，领导者可降低补贴幅度以维持相同接种率，补贴成本下降，但社会福利可能因外部性而变化；2) γ减小：追随者意愿降低，领导者需提高补贴以激励接种，成本增加。工程结论：领导者最优补贴策略与追随者偏好参数γ负相关，即γ增大时最优补贴降低。敏感性分析表明系统对γ变... | [
"基准条件复述：原模型中，追随者类型x_F服从受控马尔可夫过程，偏好参数γ影响疫苗效用项；领导者优化问题为max_{π_L} E[∑_t (W(t) - C_s(t))]，其中C_s(t)与补贴率正相关。",
"追随者最优响应：在给定π_L下，追随者均衡策略π_F*最大化J_F(x_F, μ; π_L)，疫苗选择概率p_vacc随γ增加而单调递增（因为效用增加）。",
"变化条件分析：比较γ和γ' (>γ)，在相同π_L下，追随者接种概率p_vacc(γ') > p_vacc(γ)。因此，为达到目标接种率，领导者可降低补贴。",
"多方案对比：若γ增大，领导者可降低补贴，成本C_s下降；若γ减小，领导者需提高补贴... | 2201.05959 | mimo | parameter | null | null | needs_review | {
"arxiv_id": "2201.05959",
"title": "Master Equation for Discrete Time Stackelberg Mean Field Gam",
"authors": [],
"year": null,
"paper_path": "2201.05959_Master_Equation_for_Discrete_Time_Stackelberg_Mean_Field_Gam",
"text_offset": 0
} | 考察在Stackelberg均值场博弈中，追随者偏好参数变化对领导者最优策略的敏感性，需基于模型进行多方案对比和工程结论。 | [
"未清晰列出基准和变化条件的数值对比或量化关系",
"未明确给出至少两种参数变化方向的对比分析",
"工程结论未直接关联模型或缺乏明确敏感性判定"
] | passed | The paper does not provide a specific model with a preference parameter γ affecting vaccine utility, nor does it analyze how changes in such a parameter affect the leader's optimal subsidy. The epidemic example in Section V-A describes followers' types as infection states, but no parameter γ is mentioned. The leader's ... | 0.95 | passed | passed | 0.9 | 0.8 | The paper discusses a Stackelberg mean field game with a leader and followers, and includes an epidemic model where the leader chooses subsidies for a vaccine to maximize social welfare. The answer's reasoning about the leader adjusting subsidies based on follower preferences is consistent with the paper's framework, t... | The paper's abstract and introduction describe an epidemic model where 'The leader chooses subsidies for a vaccine to maximize social welfare and minimize vaccination costs.' The answer correctly interprets this as a Stackelberg mean field game where the leader's optimal policy (subsidy) depends on follower behavior, w... | redistributable | [
"The answer introduces a preference parameter γ not present in the paper.",
"The paper does not discuss how changes in follower preferences affect the leader's optimal subsidy.",
"The answer's multi-scenario comparison and engineering conclusions are not derived from the paper."
] | [
"The answer introduces a specific parameter γ not defined in the paper excerpt, but this is a reasonable domain-specific extension for the epidemic example mentioned in the abstract.",
"The answer's conclusion that the optimal subsidy is negatively correlated with follower preference is a logical inference from t... | [
"C维缺少数值量化对比（如\"从 2.5 变为 3.8\"）"
] | null | null | null | null | [
"The answer introduces a parameter γ not explicitly defined in the provided paper excerpt, but this is acceptable as domain knowledge."
] |
CS-EVO-00033 | D | L4 | [
"nonlinear",
"robust",
"digital"
] | 针对一类反应扩散偏微分方程（PDE），设计一个性能屏障事件触发控制（P-ETC）方案，使其满足：1）全局指数收敛至零（空间L2范数意义下）；2）无Zeno现象；3）控制更新次数少于常规事件触发控制（R-ETC）。请给出完整的控制方案设计，包括数学模型、控制器结构、触发机制和参数整定方法。 | 1. 系统建模：考虑一维反应扩散PDE：u_t = α u_xx + β u + b u(1, t) + γ w(t)，空间域x∈[0,1]，边界条件u_x(0,t)=0，u(1,t)=w(t)。α,β,γ,b为已知常数。
2. 控制器选型：采用PDE反步法设计控制律w(t)=k(z(0,t))，其中k为稳定核函数，z为反步变换后的目标PDE系统状态。选择P-ETC而非R-ETC，因其允许Lyapunov函数在性能屏障V_b(t)内短暂增加，以减少触发次数。
3. 参数整定：定义性能屏障V_b(t)=V(0)e^{-λt}，λ>0。触发条件为：|V(t) - V_b(t)| ≥ εV_b(t)，ε∈(0,1)为容忍因子。控制器增益k... | [
"定义反应扩散PDE模型及其边界条件。",
"基于PDE反步法设计控制律w(t)=k(z(0,t))，并说明控制器增益k的表达式。",
"引入性能屏障V_b(t)和触发条件|V(t)-V_b(t)|≥εV_b(t)，并解释P-ETC如何允许Lyapunov函数短暂增加以减少触发。",
"证明闭环系统满足指数稳定性，并推导避免Zeno现象的最小驻留时间条件。",
"讨论工程实现中的测量、通信约束及参数调整注意事项。"
] | 2407.08178 | mimo | null | null | {
"feasibility": {
"max_score": 1,
"weight": 0.2,
"description": "方案设计是否基于论文内容，技术路线是否合理可行。"
},
"method_choice": {
"max_score": 1,
"weight": 0.2,
"description": "是否正确选择P-ETC并说明其相对于R-ETC的优势，控制方法与问题是否匹配。"
},
"completeness": {
"max_score": 1,
"weight": 0.2,
"description": "是否涵盖... | needs_review | {
"arxiv_id": "2407.08178",
"title": "Performance Barrier Event Triggered Control of a Class of Re",
"authors": [],
"year": null,
"paper_path": "2407.08178_Performance_Barrier_Event_Triggered_Control_of_a_Class_of_Re",
"text_offset": 0
} | 该题考察针对PDE系统设计P-ETC完整方案的能力，包括建模、控制律设计、触发机制、参数整定和性能分析，要求结合论文核心概念进行工程化设计。 | [
"未完整包含所有设计环节（如缺少参数整定或工程考量）。",
"数学模型或控制器表达式有误或不完整。",
"未能准确阐述P-ETC相比R-ETC的触发次数减少机制。",
"未分析避免Zeno现象的条件或性能验证不充分。"
] | rejected | The paper does not consider the PDE model u_t = α u_xx + β u + b u(1,t) + γ w(t) with boundary conditions u_x(0,t)=0, u(1,t)=w(t). Instead, it considers a reaction-diffusion PDE with boundary actuation and sensing, specifically in Section 2: 'Consider the reaction-diffusion PDE... u_t(x,t) = ε u_xx(x,t) + λ u(x,t), wit... | 0.95 | rejected | rejected | 0.95 | 0.9 | The paper does not consider the PDE model u_t = α u_xx + β u + b u(1,t) + γ w(t) with boundary conditions u_x(0,t)=0, u(1,t)=w(t). The paper's model is a reaction-diffusion PDE with boundary actuation and sensing, but the specific form and parameters given in the answer are not present. The controller structure and par... | The paper's abstract and introduction describe the P-ETC scheme for a 'class of reaction-diffusion PDEs' but do not specify the exact PDE model, boundary conditions, or controller structure given in the answer. The answer's specific model (u_t = α u_xx + β u + b u(1, t) + γ w(t)) and controller (w(t)=k(z(0,t))) are not... | redistributable | [
"The PDE model and boundary conditions do not match the paper.",
"The control law is incorrectly specified; the paper uses a backstepping controller with integral form.",
"The triggering condition is incorrect; the paper uses e(t) ≤ 0 with a dynamic variable.",
"The minimum inter-event time formula is not gen... | [
"The answer provides a specific PDE model, controller structure, and triggering condition that are not found in the provided paper excerpt. While the answer's concepts (P-ETC, backstepping, performance barrier) are consistent with the paper's domain, the detailed design is fabricated and not verifiable from the sou... | null | null | null | null | null | [
"The PDE model in the answer does not match the paper's model.",
"The controller gain formula is not from the paper.",
"The triggering condition and minimum inter-event time are not as stated in the paper.",
"The answer invents specific parameter tuning steps not present in the paper."
] |
CS-EVO-00034 | C | L4 | [
"modern",
"nonlinear"
] | 在DeepMartNet算法中，对于Dirichlet特征值问题，若将随机过程dX_t = μdt + σ·dB_t中的漂移项μ从μ₀增加到1.5μ₀，同时保持扩散项σ不变，分析由此变化对基于马尔可夫性质的损失函数设计产生的影响，并给出相应的敏感性判定。 | 基准条件：原系统参数为漂移项μ₀，扩散项σ，定义随机过程dX_t = μ₀dt + σ·dB_t，对应的椭圆算子生成器为L = μ₀ᵀ∇ + ½Tr(σσᵀ∇∇ᵀ)。变化条件：漂移项变为1.5μ₀，扩散项σ不变，新过程为dX_t = 1.5μ₀dt + σ·dB_t，新生成器L' = (1.5μ₀)ᵀ∇ + ½Tr(σσᵀ∇∇ᵀ)。变化前后的差异在于漂移项μ增加了50%，这直接影响算子L中的梯度项系数。基于马尔可夫性质的损失函数依赖于从算子L生成的随机过程X_t的采样。当μ增大时，随机过程的期望漂移速度加快，导致过程X_t在相同时间内更倾向于离开定义域D，从而影响马尔可夫鞅过程M_t = u(X_t) - u(x₀) - ∫₀ᵗ(λ... | [
"步骤1：确定基准条件下的随机过程生成器L与鞅M_t的表达式。",
"步骤2：量化分析漂移项μ从μ₀增至1.5μ₀后，对生成器L中梯度项系数的直接影响。",
"步骤3：分析该变化如何改变随机过程X_t的期望轨迹，进而影响基于马尔可夫性质的损失函数采样。",
"步骤4：对比μ增大和μ减小两种变化方向对算法采样策略和收敛性的不同影响。",
"步骤5：基于上述分析，给出明确的敏感性判定和针对工程实践的调整建议。"
] | 2307.11942 | mimo | parameter | null | null | needs_review | {
"arxiv_id": "2307.11942",
"title": "DeepMartNet A Martingale based Deep Neural Network Learni",
"authors": [],
"year": null,
"paper_path": "2307.11942_DeepMartNet_A_Martingale_based_Deep_Neural_Network_Learni",
"text_offset": 0
} | 该题考察在随机最优控制背景下，模型参数（漂移项）变化对基于马尔可夫性质的神经网络损失函数设计的影响，要求进行完整的敏感性分析、多方案对比并给出工程结论。 | [
"未明确列出变化前后的具体参数对比（如1.5μ₀）。",
"分析未紧扣损失函数（基于马尔可夫性质）的具体影响，仅泛泛而谈过程变化。",
"多方案对比不完整，仅描述了一个方向或缺少对比。",
"工程结论不明确或缺乏可操作性建议。"
] | passed | The paper does not discuss the effect of changing the drift coefficient μ on the loss function design or sensitivity analysis. The Martingale property (2.6)-(2.11) is derived for a fixed μ and σ, and no analysis of parameter variations is provided. | 0.95 | passed | passed | 0.9 | 0.9 | The paper defines the stochastic process (2.3) with drift μ and diffusion σ, and the Martingale M_t in (2.6) depends on the process X_t. The answer correctly identifies that changing μ alters the generator L and the process dynamics, affecting the sampling distribution and the Martingale property. The paper does not ex... | Section 2.1 defines the stochastic process dX_t = μdt + σ·dB_t (Eq. 2.3) and the associated elliptic operator L = μᵀ∇ + ½Tr(σσᵀ∇∇ᵀ) (Eq. 2.2). The loss function is based on the Martingale property of M_t = u(X_t) - u(x₀) - ∫₀ᵗ(λ - V(X_s))u(X_s)ds (Eq. 2.6), whose expectation is constant (Eq. 2.11). The answer correctly... | redistributable | [
"The answer speculates on the impact of changing μ without any support from the paper.",
"The paper does not mention sensitivity analysis or adaptive sampling strategies.",
"The answer introduces concepts (e.g., 'sampling time step adjustment', 'adaptive sampling') not present in the source."
] | [
"Minor: The answer's discussion of sensitivity and engineering recommendations (e.g., adjusting sampling time steps) is reasonable domain knowledge but goes beyond the explicit content of the provided paper excerpt, which does not analyze parameter sensitivity or propose adaptive strategies."
] | [
"C维缺少数值量化对比（如\"从 2.5 变为 3.8\"）"
] | null | null | null | null | null |
CS-EVO-00035 | D | L4 | [
"classical",
"digital"
] | 针对论文中描述的重型四旋翼无人机在移动船只上降落的场景，设计一个完整的PID姿态控制系统（俯仰通道），以确保无人机在遭遇垂直振荡和横风干扰时，能将俯仰角跟踪误差控制在±2°以内，且调节时间小于1秒。 | 1. 系统建模：基于论文中描述的四旋翼动力学，俯仰通道简化模型为：Jθ θ̈ = uθ，其中 Jθ 为俯仰转动惯量，uθ 为俯仰力矩控制输入。考虑风扰动 d(t) 和船体垂直振荡引起的姿态误差 e_d(t)，则模型为 Jθ θ̈ = uθ - Jθ d(t) - Jθ e_d(t)。
2. 控制器选型：选用PID控制器，因为其结构简单、实时性好，适用于论文中提到的实时控制系统，且能通过积分项消除常值扰动。控制律为 uθ = Kp(θ_ref - θ) + Ki ∫(θ_ref - θ)dt + Kd(θ̇_ref - θ̇)。
3. 参数整定：基于文献中类似模型参数（如 Jθ = 0.01 kg·m²），采用Ziegler-Nich... | [
"基于论文描述，建立俯仰通道简化动力学模型，包含转动惯量、控制力矩和扰动项。",
"选用PID控制器，因其适用于论文中的实时控制场景，并能处理积分误差。",
"采用Ziegler-Nichols方法整定PID参数，给出具体数值和理论依据。",
"通过传递函数分析闭环系统性能，包括稳定裕度、响应时间和误差指标。",
"讨论实际实现中的约束，如执行器饱和、传感器噪声和计算频率要求。"
] | 2411.05445 | mimo | null | null | {
"feasibility": {
"max_score": 1,
"weight": 0.2,
"description": "方案是否基于论文描述的无人机和场景，技术路径是否可行。"
},
"method_choice": {
"max_score": 1,
"weight": 0.2,
"description": "控制器选型是否合理，理由是否充分，是否匹配控制目标。"
},
"completeness": {
"max_score": 1,
"weight": 0.2,
"description": "是否覆盖建模、选型、整定、验... | needs_review | {
"arxiv_id": "2411.05445",
"title": "Agile UAV landing control on moving ship in adverse conditio",
"authors": [],
"year": null,
"paper_path": "2411.05445_Agile_UAV_landing_control_on_moving_ship_in_adverse_conditio",
"text_offset": 0
} | 该题考察针对特定工程场景（移动平台上的无人机降落）设计完整姿态控制方案的能力，包括建模、控制器设计、参数整定、性能分析和工程实现考量，要求方案具体、参数可验证、分析全面。 | [
"建模部分若遗漏扰动项或简化过度，可能导致方案不完整。",
"控制器选型理由不充分或未与论文场景关联。",
"参数整定过程缺失或数值不合理，未基于理论方法。",
"性能分析未覆盖所有关键指标（如稳定裕度、响应时间、稳态误差）或计算错误。",
"工程考量未涉及执行器限制、传感器处理或实时性等实际约束。"
] | passed | The paper does not provide specific quadrotor dynamics equations, PID controller design, or parameter tuning details. It only mentions using a PID flight control system based on reference-point tracking, but no explicit model or controller parameters are given. The answer invents numerical values (Jθ=0.01, Ku=10, Tu=0.... | 0.95 | passed | passed | 0.9 | 0.8 | The paper describes a PID flight control system based on reference-point tracking for a quadrotor UAV, including realistic saturation limits and disturbance handling. The answer designs a PID controller for the pitch channel, which is consistent with the paper's approach. The paper does not provide specific PID gains o... | The paper states: 'The PID flight control system is proposed based on reference-point tracking, allowing the UAV to follow any desired path in 3D space whilst simultaneously yawing to face any desired heading.' and 'Realistic saturation limits on actuator outputs to ensure the real-world performance of actuators.' The ... | redistributable | [
"The paper does not provide a simplified pitch dynamics model Jθ θ̈ = uθ.",
"No specific PID gains or tuning method (Ziegler-Nichols) are mentioned.",
"The numerical values for Jθ, Ku, Tu, Kp, Ki, Kd are fabricated.",
"The closed-loop transfer function and performance metrics (phase margin, gain margin, settl... | [
"The answer's simplified model (Jθ θ̈ = uθ) omits the coupling effects and nonlinear dynamics typical of a quadrotor, which the paper implies by discussing a full 3D control system. However, for a single-channel design question, this simplification is a reasonable engineering approach and does not contradict the pa... | null | null | null | null | null | null |
CS-EVO-00036 | B | L3 | [
"classical",
"robust"
] | 考虑一个简化的单电池模组，其状态空间模型为：dx/dt = Ax + Bu，y = Cx。其中状态x = [SOC, T]^T（SOC为荷电状态，T为温度），输入u为充电电流I_ch，参数矩阵为A = [[-0.01, 0.001], [0.005, -0.02]]，B = [[0.002], [0.001]]，C = [[1, 0]]。假设系统初始状态x0 = [0.8; 25]，在t=0时刻，攻击者通过篡改BMS传感器读数，使得系统实际接收到的输出测量值y_meas = y + a(t)，其中a(t) = 0.1（常数偏移攻击）。若BMS基于该受污染的测量值y_meas，采用状态反馈控制律u = -K*x_hat，且设计增益K使... | 系统矩阵A的特征值为λ1≈-0.013，λ2≈-0.017，系统开环稳定。参考值y_ref=0.8，对应开环稳态输入u_ref满足0 = A*x_ss + B*u_ref，y_ref = C*x_ss，解得x_ss = [0.8; 24.75]，u_ref = -0.01。攻击者篡改输出y_meas = y + 0.1，BMS基于y_meas估计状态x_hat并实施反馈u = -K*x_hat，闭环期望极点-0.1，-0.2，可解得K。为使输出y(t)跟踪y_ref=0.8，攻击者需施加输入攻击u_a(t)，使系统状态轨迹满足dx/dt = A*x + B*(u_ref + u_a(t))，且输出y = C*x = 0.8恒成立。将... | [
"第一步：基于给定参数矩阵A、B、C和参考输出y_ref=0.8，计算开环稳态状态x_ss和输入u_ref。由0=A*x_ss+B*u_ref和y_ref=C*x_ss，解得x_ss=[0.8; 24.75]，u_ref=-0.01。",
"第二步：分析隐蔽攻击条件。攻击者通过输出偏移a(t)=0.1污染测量值y_meas，为使实际输出y(t)=C*x(t)仍跟踪y_ref=0.8，需满足C*x(t)=0.8恒成立，即SOC恒为0.8。",
"第三步：将SOC恒为0.8的条件代入状态方程第一行：d(SOC)/dt = A(1,1)*SOC + A(1,2)*T + B(1)*(u+u_a(t)) = 0。已知u采用基... | 2401.13801 | mimo | null | null | null | needs_review | {
"arxiv_id": "2401.13801",
"title": "Exploring Adversarial Threat Models in Cyber Physical Batter",
"authors": [],
"year": null,
"paper_path": "2401.13801_Exploring_Adversarial_Threat_Models_in_Cyber_Physical_Batter",
"text_offset": 0
} | 该题考察从网络化控制系统（CPBS）的对抗性威胁模型出发，结合状态空间表示、反馈控制设计及隐蔽攻击条件，进行多步数学推导以推导攻击信号表达式的能力，综合应用了状态反馈控制、系统建模和对抗分析。 | [
"未能正确建立或求解开环稳态条件，导致初始推导错误。",
"未能正确理解并应用隐蔽攻击条件（输出恒为参考值），导致推导方向偏离。",
"在推导u_a(t)表达式时，未能清晰隔离出攻击者可控部分并给出最终结果。",
"计算过程有误，导致最终数值结果错误。"
] | passed | The source paper does not provide any specific state-space model, parameter matrices, or control design details. The paper focuses on adversarial threat models for battery systems but does not include the numerical example or derivations presented in the answer. Therefore, the answer is not factually supported by the p... | 0.95 | passed | passed | 0.9 | 0.9 | The paper discusses adversarial attack generation in battery systems, focusing on stealthy attacks that manipulate sensor readings while remaining undetected. The answer describes a scenario where an attacker injects a constant offset into the output measurement and derives an input attack signal to maintain the output... | The paper discusses adversarial threat models for cyber physical battery systems, focusing on attack generation policies and stealthiness. The answer correctly applies control theory concepts (state-space models, feedback control, attack signals) consistent with the paper's domain of analyzing adversarial dynamics in b... | arxiv-only | [
"The answer invents a specific state-space model and parameters not present in the paper.",
"The derivation of input attack signal u_a(t) is not based on any content from the paper.",
"The paper does not discuss state feedback control design or pole placement for battery systems."
] | [
"Minor: The answer's numerical calculations (e.g., eigenvalues, steady-state values) are not verified against the paper, but this is acceptable as the paper does not provide specific numerical examples for this scenario."
] | null | null | null | null | null | null |
End of preview. Expand in Data Studio
Sci-Align Benchmark
1,738 control-science questions with reference answers and written audit reasoning, from the Sci-Align Benchmark evaluation project. Every item carries a machine-readable licence bucket.
This mirror is the project's question pool with one bucket removed: items whose source is
restricted (non-commercial / no-derivatives terms) are excluded here
(241 items). The full 1,979-item pool, including those items,
remains in the GitHub repository.
Load it
from datasets import load_dataset
ds = load_dataset("MorningStar0709/sci-align-benchmark", split="train") # 1,738 rows
The single data file is core_v3_hf.json, a JSON array of question objects. How this mirror was
cut from the upstream pool is recorded below. source_pool_questions_sha256 is the SHA-256 of that
pool's questions array, JSON-dumped with sorted keys and compact separators
(json.dumps(qs, ensure_ascii=False, sort_keys=True, separators=(",", ":"))) — it does not move when
the pool's metadata is rebuilt, only when its questions change:
{
"license_scope": "A",
"excluded_buckets": [
"restricted"
],
"excluded_items": 241,
"kept_items": 1738,
"counts_by_bucket": {
"redistributable": 685,
"arxiv-only": 927,
"textbook-unchecked": 126
},
"source_pool": "benchmark/dataset/core_v3.json",
"source_pool_questions_sha256": "fefc7ab7fe0912b476ced08841d54ffc00e436fccc470c8b9b8ec31c0674f0b2"
}
What is in each item
question, answer, reasoning_steps, rubric, dimensions and difficulty, plus audit fields —
per-item scoring_reason, audit_* and their evidence notes. Questions and reference answers are
written by us; sources are cited by ID (source_ref), never reproduced wholesale.
Licence buckets
| bucket | meaning | items |
|---|---|---|
redistributable |
redistributable (CC BY etc.) | 685 |
arxiv-only |
distributed via arXiv only | 927 |
textbook-unchecked |
textbook source; no per-item determination | 126 |
redistributable— the source is under a licence that permits redistribution with attribution.arxiv-only— the source is distributed through arXiv; that is not a grant to redistribute derivatives. Items here cite the source at short length and are labelled as such.textbook-unchecked— the source is a textbook with no per-source licence determination. Stated here, not hidden.
Where an audit note quotes a source to justify a judgement, verbatim spans of 300 characters or
more are replaced by a placeholder; our own wording and shorter quotations stay. The rule, its
counts and the per-source breakdown are in the repository's LICENSE-LEDGER.md and SOURCES.md.
Licence and takedown
The questions and reference answers we wrote are CC-BY-4.0; third-party sources keep their own
terms, which is why the bucket column above is per item. If you hold rights to any item's source and
want it removed, open an issue on the repository naming the item IDs (CS-EVO-#####), and it will be
removed from this mirror.
Citation
Metadata — authors included — is in the repository's CITATION.cff. The project's two preprints:
10.5281/zenodo.23409480 (protocol) and
10.5281/zenodo.23409394 (findings).
- Downloads last month
- 27