面向空间航天器机器人学习算法研究的高保真仿真平台

A High-Fidelity Simulation Environment for Spacecrafts with Robot Learning Algorithms

针对空间航天器失重服役环境下样本数据少、实验测试难等问题,构建了一个面向空间航天器的机器人学习算法研究的高保真仿真平台.该平台依照机器人学习领域标准的数据交互模式,支持直接调用各类主流机器人学习算法库,以及按照Gym交互模式开发的控制/学习算法;利用微重力模拟系统的实验数据,采用数据驱动的方式构建航天器动力学模型,用于仿真平台中的状态更新;以航天器位置姿态稳定任务为例,在所构建的仿真平台中完成了主流强化学习算法Soft Actor-Critic的训练和测试,验证了所构建的仿真平台用于机器人学习算法研究的可行性.

Robot learning algorithms have promoted the development of motion planning and control,and a key is-sue in robot learning is how to building a high-performance robot physics engine.Due to the special environment of spacecrafts,characterized by limited sample data and costly experimental conditions,this paper presents a high-fi-delity simulation environment for spacecrafts with robot learning algorithms.Adhering to the standard Gym frame-work,the simulation environment supports a variety of mainstream robot learning algorithm libraries and Gym-style control/learning algorithms.Utilizing experimental data from the microgravity simulation system,a data-driven ap-proach is employed to construct a spacecraft dynamics model for state updates within the simulation environment.As an illustrative example,the mainstream reinforcement learning algorithm Soft Actor-Critic is trained and tested in the constructed simulation environment for the spacecraft stabilization task,demonstrating the feasibility of the sim-ulation environment for robot learning algorithm.