GPUDrive：基于数据驱动的，每秒100万帧的多智能体驾驶模拟。

GPUDrive: Data-driven, multi-agent driving simulation at 1 million FPS

August 2, 2024

作者: Saman Kazemkhani, Aarav Pandya, Daphne Cornelisse, Brennan Shacklett, Eugene Vinitsky

cs.AI

摘要

多智能体学习算法在各种游戏中成功生成了超人类规划，但对部署的多智能体规划器设计影响甚微。将这些技术应用于多智能体规划的一个关键瓶颈是它们需要数十亿步的经验。为了实现对这一规模的多智能体规划的研究，我们提出了GPUDrive，这是一个基于Madrona游戏引擎构建的GPU加速多智能体模拟器，每秒可以生成超过一百万步的经验。观察、奖励和动力学函数直接用C++编写，使用户能够定义复杂、异构的智能体行为，并转换为高性能的CUDA。我们展示了使用GPUDrive，我们能够在Waymo Motion数据集的许多场景中有效训练强化学习智能体，在几分钟内为单个场景生成高效的目标达成智能体，通常在几小时内生成具备一般能力的智能体。我们将这些训练有素的智能体作为代码库的一部分发布在https://github.com/Emerge-Lab/gpudrive。

English

Multi-agent learning algorithms have been successful at generating superhuman planning in a wide variety of games but have had little impact on the design of deployed multi-agent planners. A key bottleneck in applying these techniques to multi-agent planning is that they require billions of steps of experience. To enable the study of multi-agent planning at this scale, we present GPUDrive, a GPU-accelerated, multi-agent simulator built on top of the Madrona Game Engine that can generate over a million steps of experience per second. Observation, reward, and dynamics functions are written directly in C++, allowing users to define complex, heterogeneous agent behaviors that are lowered to high-performance CUDA. We show that using GPUDrive we are able to effectively train reinforcement learning agents over many scenes in the Waymo Motion dataset, yielding highly effective goal-reaching agents in minutes for individual scenes and generally capable agents in a few hours. We ship these trained agents as part of the code base at https://github.com/Emerge-Lab/gpudrive.