Robotic Control Benchmarks
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/HcPlu/Evolutionary-Constrained-Reinforcement-Learning
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了一系列连续的机器人控制任务，其中包括具有扭矩限制的Ant、HalfCheetah、Walker2d、Hopper和Swimmer等任务。这些任务主要关注在特定的扭矩限制下最大化行走性能，并在文献中被广泛使用。数据集的规模涉及多个任务，具有不同观测和动作维度。总体任务目标是，在机器人控制中，在扭矩限制下最大化奖励。
This dataset encompasses a series of continuous robotic control tasks, including Ant, HalfCheetah, Walker2d, Hopper, and Swimmer with torque constraints. These tasks primarily focus on maximizing walking performance under specified torque constraints, and have been widely adopted in academic literature. The dataset covers multiple tasks with distinct observation and action dimensions. The overall task objective is to maximize the reward under torque constraints in robotic control scenarios.
提供机构：
OpenAI Gym
