IntegratedMotionPrediction

Integrated Data-driven Inference and Planning-based Human Motion Prediction for Safe Human-Robot Interaction

Overall Architecture

We develop the algorithm for a autonomous vehicle to safely and actively interact with an uncertain human-driven vehicle.

The algorithm combines:

• A hierarchical prediction strategy that integrates data-driven human internal state inference with planning-based human motion prediction
• An Active motion planning algorithm for the autonomous vehicle to ensure safety against uncertain human motions

Conference Proceeding

Y. Nam and C. Kwon, “Integrated Data-driven Inference and Planning-based Human Motion Prediction for Safe Human-Robot Interaction”, ICRA 2024: International Conference on Robotics and Automation, Yokohama, Japan, May 2024

How to train inference module

1. Generate human driving data for training
   -> python ./gen_human_data.py
   
2. Train rationality inference module
   -> python ./train_inference/train_beta.py
   
3. Train driving style inference module
   -> python ./train_inference/train_psi.py
   

How to run main algorithm

1. Run main code
   -> python ./main.py
   ** After training the inference module, 'model_id' for each inference module should be changed accordingly.