GUIDE-CoT

The official implementation code for "GUIDE-CoT: Goal-driven and User-Informed Dynamic Estimation for Pedestrian Trajectory using Chain-of-Thought" [AAMAS 2025]

Acknowledgements

This project significantly builds upon the work of others. We extend our sincere gratitude to the authors and developers of the following projects:

• The goal_module is largely based on the implementation of Goal-SAR. We express our sincere gratitude for their work and for making their code publicly available.

• The llm_module incorporates and adapts significant portions of code from LMTraj-SUP. We deeply appreciate their contributions and the accessibility of their resources.

Installation

• All experiments were performed in an Ubuntu 20.04, Python 3.9, RTX 3090Ti environment.

• Install the following conda environment.

  $ conda env create -f guide-cot.yaml
  $ conda activate GUIDE-CoT

• Install OpenAI CLIP Library following this [link]

Preprocess

$ bash scripts/preprocess_all.sh

Training

$ bash scripts/train_all.sh

Evaluation

$ bash scripts/test_all.sh

Citation

@inproceedings{kim2025guide,
  title={GUIDE-CoT: Goal-driven and User-Informed Dynamic Estimation for Pedestrian Trajectory using Chain-of-Thought},
  author={Kim, Sungsik and Baek, Janghyun and Kim, Jinkyu and Lee, Jaekoo},
  booktitle={Proc. of the 24th International Conference on Autonomous Agents and Multiagent Systems},
  pages={1107--1116},
  year={2025}
}