MOTIONBENCHMAKER

MOTIONBENCHMAKER是一个由莱斯大学开发的开源工具，用于生成和基准测试机器人运动规划数据集。该数据集包含40个预制数据集，涵盖5种常用机器人和8种环境，旨在为运动规划研究提供一个公平的评估平台。数据集通过随机生成场景和参数来创建，支持将场景转换为感知表示，如点云和八叉图，并兼容ROS生态系统。MOTIONBENCHMAKER的应用领域广泛，包括学习型规划方法、超参数调整、不确定性下的规划以及部分可观测环境中的规划，旨在解决运动规划中的复杂问题。

MOTIONBENCHMAKER is an open-source tool developed by Rice University for generating and benchmarking robotic motion planning datasets. This toolkit includes 40 pre-built datasets covering 5 commonly used robots and 8 types of environments, aiming to provide a fair evaluation platform for motion planning research. The datasets are created via randomly generating scenarios and parameters, support converting scenarios into perceptual representations such as point clouds and octomaps, and are compatible with the ROS ecosystem. MOTIONBENCHMAKER has a wide range of application scenarios, including learning-based planning methods, hyperparameter tuning, planning under uncertainty, and planning in partially observable environments, and is designed to address complex issues in motion planning.