iPlanner weights (mirror)

Mirror of the pre-trained plannernet.pt of iPlanner: Imperative Path Planning (Yang, Wang, Cadena and Hutter, RSS 2023, arXiv 2302.11434), for the iplanner planner of arena_planners.

• Source: https://github.com/leggedrobotics/iPlanner (commit 4a8d823, 2025-02-23), Google Drive file 1UD11sSlOZlZhzij2gG_OmxbBN4WxVsO_ linked from the README under "Pre-trained Network and Training Data".
• License: MIT, the upstream repository license (Copyright (c) 2023 Fan Yang, Robotic Systems Lab, ETH Zurich). The README states "This code is released under the MIT License" and attaches no other terms to the weights. Upstream notes the network was trained in simulation and not adapted to real-world data.

Files

File	sha256	Content
plannernet.pt	685f16cde28d05249d50d24ed79ab4bdc94b3fbbcb99c8dbaed31039d11633b9	the upstream file unchanged, a pickled (PlannerNet, best_loss) tuple that needs the upstream planner_net and percept_net modules and weights_only=False to load
plannernet_state_dict.pt	beccb331e782130b4ebe7738020af89104a94e993f8a95f1b5dfec1a9ebf6cb4	the same tensors as a plain state_dict (51 keys, 53,326,864 parameters), loads with torch.load(weights_only=True) into PlannerNet(encoder_channel=16, k=5)
convert.py		the script that produced the state_dict file, run with the vendored iplanner_model package on the path

Citation

@inproceedings{Yang-RSS-23,
  author    = {Fan Yang and Chen Wang and Cesar Cadena and Marco Hutter},
  title     = {{iPlanner: Imperative Path Planning}},
  booktitle = {Proceedings of Robotics: Science and Systems},
  year      = {2023},
  address   = {Daegu, Republic of Korea},
  month     = {July},
  doi       = {10.15607/RSS.2023.XIX.064}
}