MultiCarRacing-v0
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/igilitschenski/multi_car_racing
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是一个创新的多智能体赛车环境，用于学习具有竞争力的视觉控制策略。它基于Gym任务CarRacing-v0进行扩展，并为每个智能体提供从自我视角的俯视96x96像素图像观察。该环境能够区分熟练驾驶与竞争性驾驶，包含复杂的动态效果，如滑行和碰撞。在此规模上，多个智能体在赛道上进行比赛，任务是在多智能体赛车设置中学习竞争性行为。
This dataset is an innovative multi-agent racing environment developed for learning competitive visual control policies. It is extended from the OpenAI Gym benchmark task CarRacing-v0, and provides each agent with a top-down 96×96 pixel visual observation from its own egocentric perspective. This environment can distinguish between proficient driving and competitive racing, and includes complex dynamic behaviors such as skidding and collisions. Multiple agents compete on the race track in this environment, with the task of learning competitive driving behaviors within the multi-agent racing setup.
