Scale-OSPO

Article: Zijian Zhao, Sen Li*, "OSPO: One-Step Policy Optimization for Real-Time Order Assignment on Ride-Sharing Platforms with Autonomous Vehicles" (under review)

Please refer to RS2002/OSPO: Official Repository for The Paper, One Step is Enough: Multi-Agent Reinforcement Learning based on One-Step Policy Optimization for Order Dispatch on Ride-Sharing Platforms.

Execute the following command in the ./OSPO/pretrain directory:

python train.py

Execute the following command in the ./OSPO/finetune directory:

python train.py --model_path <path_to_trained_model_from_pretraining>