SafeFlowMPC
收藏资源简介:
SafeFlowMPC微调数据集是为机器人安全轨迹规划研究而设计的数据集,主要用于论文《SafeFlowMPC: Predictive and Safe Trajectory Planning for Robot Manipulators with Learning-based Policies》中描述的微调阶段。该数据集结合了流匹配和在线优化方法,旨在实现安全灵活的机器人操作。数据集包含通过VP-STO规划器生成的安全中间轨迹,用于训练流匹配模型执行精确且安全的机器人任务,如抓取和人机物体交接。数据集包含199,099个训练样本,总大小约为187.8GB,特征包括轨迹数据(trajectories)、控制数据(c_data)、样本数据(samples)及其衍生数据(dsamples, t_samples)等多维浮点数组。该数据集专为与KUKA 7自由度机械臂相关的任务设计,需配合官方训练脚本使用。
The SafeFlowMPC fine-tuning dataset is designed for research on robotic safe trajectory planning, and is primarily used for the fine-tuning stage described in the paper *SafeFlowMPC: Predictive and Safe Trajectory Planning for Robot Manipulators with Learning-based Policies*. This dataset combines flow matching and online optimization methods, aiming to achieve safe and flexible robotic manipulation. It contains safe intermediate trajectories generated by the VP-STO planner, which are utilized to train flow matching models to perform precise and safe robotic tasks such as grasping and human-robot object handover. The dataset comprises 199,099 training samples with a total size of approximately 187.8 GB, and its features include multi-dimensional floating-point arrays such as trajectory data (trajectories), control data (c_data), sample data (samples), and their derived data (dsamples, t_samples). This dataset is specifically tailored for tasks related to KUKA 7-degree-of-freedom robotic arms, and must be used in conjunction with the official training scripts.



