RL-CIS
收藏资源简介:
RL-CIS是中国科学院自动化研究所创建的数据集,专注于自动驾驶在复杂交叉路口场景的训练与测试。该数据集包含5个功能场景,涵盖左转、右转和直行等典型驾驶任务,通过CARLA模拟器生成,旨在通过强化学习方法提升自动驾驶系统的决策与控制能力。数据集的创建过程中,采用了基于随机过程的交通流生成方法,确保场景的多样性和随机性。RL-CIS主要应用于自动驾驶系统的性能评估,特别是在处理与社会车辆交互的复杂情况时,以提高交通安全和效率。
RL-CIS is a dataset developed by the Institute of Automation, Chinese Academy of Sciences, focusing on the training and testing of autonomous driving systems in complex intersection scenarios. This dataset includes 5 functional scenarios, covering typical driving tasks such as left turn, right turn, and straight driving, and is generated via the CARLA simulator. It aims to enhance the decision-making and control capabilities of autonomous driving systems through reinforcement learning approaches. During the dataset's development, a traffic flow generation method based on stochastic processes was adopted to ensure the diversity and randomness of the scenarios. RL-CIS is primarily utilized for the performance evaluation of autonomous driving systems, especially in complex situations involving interactions with social vehicles, to improve traffic safety and efficiency.

- 1A Reinforcement Learning Benchmark for Autonomous Driving in Intersection Scenarios中国科学院自动化研究所 · 2021年



