핵심 개념
Addressing the covariate shift problem in multi-agent imitation learning for realistic traffic simulation.
통계
"achieving 40x simulation length improvements over previous state-of-the-art"
"dataset pNEUMA with over half a million trajectories"
"our method achieves better results than all baselines in terms of position and velocity RMSE, road density and speed RMSE, with minor off-road rate"
인용구
"Our method has demonstrated superior performance over existing state-of-the-art simulators"
"Augmentation plays a vital role in improving both short-term and long-term simulation performance"