This repository contains the code for the papers published with the Risk Aware Driving (RAD) team at Toyota Research Institute (TRI), including, but not limited to: Xin Huang, Rui Yu, Guy Rosman, Deepak Gopinath, Jon Decastro, Stephen G. McGill, Igor Gilitschenski, Xiongyi Cui, Yen-Ling Kuo, Thomas Balch, Paul Drews, Mark Flanagan, Caleb Severn, Nicholas Guyett, Ankur Kalra, Sarah Andrews, Justin Lidard.
TRI would like to acknowledge and recognize the help of Hop Labs team members on this and other projects.
This repo contains the following packages and folders:
data_sources- Converters for different data sources.intent- Training and experiment scripts.model_zoo- The pytorch models.radutils- RAD research utilities.triceps- TRI Common Environment and Prediction Serializer.
See prediction_framework_overview.pdf for additional details.
See below in order to run the code for each paper under intent/multiagents/:
hybrid- Code for "HYPER: Learned Hybrid Trajectory Prediction via Factored Inference and Adaptive Sampling", paper by: Xin Huang, Guy Rosman, Igor Gilitschenski and Ashkan Jasour and Stephen G. McGill, John J. Leonard and Brian C. Williams.language- Code for "Trajectory Prediction with Linguistic Representations", paper by Yen-Ling Kuo, Xin Huang, Andrei Barbu, Stephen G. McGill, and Boris Katz and John J. Leonard and Guy Rosman.
Conda is used to create the development environment.
- To create the environment using existing conda.
conda env create -f environment.pt190.ymlOR
- Let it download miniconda to ~/miniconda3
./scripts/create_env.sh --skip-awsThen
Activate the conda environment, before running the code.
conda activate pt190Toyota Research Institute would like to acknowledge and recognize the work of the Hop Labs team members on this and related projects: Mark Flanagan, Caleb Severn, Nicholas Guyett, Sarah Andrews and Ankur Kalra.
See the LICENSE.md file for details.
Copyright 2018-2022 TRI.