Continental-Empirical-Bayes-Analysis

Demo for extracting prior knowledge (model parameter distribution) from Waymo Open Motion Dataset v1.1 via Empirical Bayes Analysis.

Getting started

Installing

1. Clone this repository
2. Download Waymo Open Motion Dataset and put the data into "demo/data" directory
3. Install Dependencies
   1. pip install requirements.txt
   2. install waymo_open_dataset according to here

Running

1. Run demo/preprocess_waymo.ipynb as entrypoint.
2. Excute Empirical Bayes Analysis with:
   1. demo/ego_traj_analyse.ipynb for ego vehicle trajectory.
   2. demo/agt_traj_analyse.ipynb for other object trajectory.
3. Evaluate AIC, BIC and representation error with demo/result_evaluation.ipynb.

Prior knowledge

• We provide the extracted model prior distribution and observation noise distribution for different objects and timescales in demo/logs/gradient_tape.
• An example of integrating prior knowledge is provided in demo/integrate_prior.ipynb.

Reference

1. An Empirical Bayes Analysis of Object Trajectory Models

Documentation

For further help, see the API-documentation or contact the maintainers.

License

Copyright (c) 2023 Continental Corporation. All rights reserved.