SDM (Stackelberg Driver Model) (https://arxiv.org/abs/2309.14235) is a novel approach that aligns with the hierarchical nature of interactions between autonomous vehicle (AV) and background vehicles (BVs), facilitating AV performance improvement in a leader-follower order of play. We empirically demonstrate that AV and BV agents can both achieve better performance during the game with SDM. Essentially, SDM outperforms those competing baselines with simple adversarial training paradigms, non-game modeling and other game modeling approaches, especially in higher dimensional scenarios.
To install the dependencies, run the command:
conda create --name SDM --file requirements.txt
We run experiments on SDM, Non-Game, SimGM, NSD and I-SDM. To begin with, enter the folder Scripts
:
cd Scripts
Then you can run SDM and baselines using following example commands.
python main_spg_ac_inv.py --r_ego stackelberg --r_adv stackelberg --is_save True --save_model True --adv_policy sumo --num_agents 5
python main_NSG.py --r_adv stackelberg3 --r_ego stackelberg --is_save True --save_model True --pretrain_ego True --adv_policy sumo --num_agents 5
python main_SimGM.py --r_adv stackelberg2 --r_ego stackelberg --is_save True --save_model True --pretrain_ego True --adv_policy sumo --num_agents 5
You can resort to wandb to login your personal account with your wandb API key.
export WANDB_API_KEY=YOUR_WANDB_API_KEY
and using --used_wandb True
to turn on the online syncronization.
If you are using SDM framework or code for your project development, please cite the following paper:
@article{niu2023stackelberg,
title={Stackelberg Driver Model for Continual Policy Improvement in Scenario-Based Closed-Loop Autonomous Driving},
author={Niu, Haoyi and Chen, Qimao and Li, Yingyue and Hu, Jianming},
journal={arXiv preprint arXiv:2309.14235},
year={2023}
}