OPGP

This repo is the implementation of:

Occupancy Prediction-Guided Neural Planner for Autonomous Driving
Haochen Liu, Zhiyu Huang, Chen Lv
AutoMan Research Lab, Nanyang Technological University
[arXiv] [Zhihu]

Code is now released 😀!

Overview

In this repository, you can expect to find the following features 🤩:

Pipelines for data process and training
Open-loop evaluations

Not included 😵:

Model weights (Due to license from WOMD)
Real-time planning (Codes are not optimized for real-time performance)

Experiment Pipelines

Dataset and Environment

Downloading Waymo Open Motion Dataset v1.1. Utilize data from scenario/training_20s for train set, and data from scenario/validation for val & test.
Clone this repository and install required packages.
[NOTED] For theseus library, you may build from scratch and add system PATH in planner.py

Data Process

Preprocess data for training & testing:

python preprocess.py \
--root_dir path/to/your/Waymo_Dataset/scenario/ \
--save_dir path/to/your/processed_data/ \
--processes=16

You may also refer to Waymo_candid_list for more interactive and safety-critical scenarios filtered in scenario/validation

Training & Testing

Train & Eval the model using the command:

python -m torch.distributed.launch \
        --nproc_per_node 1 \ # number of gpus
        --master_port 16666 \
        training.py \
        --data_dir path/to/your/processed_data/ \
        --save_dir path/to/save/your/logs/

Conduct Open-loop Testing using the command:

python testing.py \
        --data_dir path/to/your/testing_data/ \
        --model_dir path/to/pretrained/model/

Citation

If you find this repository useful for your research, please consider giving us a star 🌟 and citing our paper.

@article{liu2023occupancy,
  title={Occupancy Prediction-Guided Neural Planner for Autonomous Driving},
  author={Liu, Haochen and Huang, Zhiyu and Lv, Chen},
  journal={arXiv preprint arXiv:2305.03303},
  year={2023}
}

jyakaranda / OPGP