Update README.md

README.md

@@ -1,3 +1,55 @@
 ---
 license: mit
+tags:
+- unity
+- ragdoll
+- onnx
+- machine learning
+- active ragdoll
+- vision
+- lstm
 ---
+
+# Kyle - Robot Walker Ragdoll Trainer
+
+## Model Description
+Kyle is a state-of-the-art active ragdoll training environment in Unity, featuring a heavily modified and optimized codebase from the original Ragdoll Trainer. Our version includes advanced vision capabilities and short-term variable memory using LSTM (Long Short-Term Memory) networks, providing a more sophisticated simulation for AI training.
+
+## How to Use the Model
+To ensure proper functionality of the Ragdoll Trainer within the Neko Cat Game, it is essential to utilize the **skeletal armature GameObject** provided with our modified project. This GameObject is specifically tailored for our Ragdoll Trainer and can be found as a child object within the main project directory at `https://github.com/cat-game-research/Neko/tree/main/RagdollTrainer`.
+
+1. Import the ONNX model into your Unity project.
+2. Attach the model to the skeletal armature GameObject included with our modified Ragdoll Trainer project.
+3. Utilize the model's capabilities to train and simulate realistic ragdoll movements with enhanced cognitive functions.
+
+## Training Data
+The model was trained on a comprehensive dataset that includes various ragdoll movements, enriched with visual inputs and memory components to simulate real-world scenarios effectively.
+
+## Evaluation Results
+Kyle has undergone extensive testing, showcasing remarkable proficiency in simulating lifelike ragdoll physics with the added depth of vision and memory-based learning.
+
+## Ethical Considerations
+We adhere to ethical AI development practices, ensuring our models are created with fairness, accountability, and transparency at their core.
+
+## Acknowledgements
+We extend our gratitude to the creators of the original Ragdoll Trainer project for their foundational work. Our enhancements were made possible by building upon their innovative approach to active ragdoll training.
+
+## Additional Information
+- **Author**: p3nGu1nZz
+- **Unity Version**: 2023.2.7f1
+- **ONNX Version**: 1.12.0
+- **ML-Agents Version**: 3.0.0-exp.1
+- **ML-Agents Extensions Version**: 0.6.1-preview
+- **Sentis Version**: 1.3.0-pre.2
+- **Features**:
+  - Enhanced rig created in Blender for more natural movement.
+  - Heuristic function for joint control during development testing.
+  - Stabilizers for hips and spine to aid balance.
+  - Early training settings for initial balance and walking towards targets.
+  - Navigation around obstacles using Ray Perception Sensor 3D.
+  - Training for navigating steps and stairs with varying difficulty.
+- **Setup Process**:
+  - Local Miniconda installation and configuration.
+  - Cloning of the latest tag of ML-Agents into the project.
+  - Script configuration for fine-tuning models based on Kyle.
+