摘要
以小行星表面着陆探测为背景,提出一种动量驱动机器人(MoRo)以满足弱引力复杂环境下的探测需求。该机器人利用弱引力环境下的摩擦和碰撞特性,通过主动辨识环境参数,规划和控制动量轮以产生期望的驱动力矩,完成可控性跳跃及腾空后的稳定拍照等任务。首先,基于MoRo的动量轮刹车机构特性,分析了MoRo在弱引力环境下的跳跃机理并对其跳跃方式进行了规划;接着考虑动量轮驱动机构三闭环伺服系统的非线性特性,基于Herze碰撞模型和Karnopp摩擦模型建立了MoRo在小行星表面的跳跃行为动力学模型;其次,使用机器学习算法建立环境参数和MoRo运动的函数关系,并基于环境参数规划动量轮转速实现跳跃距离和腾空高度的可控。最后,通过数值仿真校验了MoRo跳跃规划方法和控制方法的可行性。
To meet the detection requirement in a weak complex gravitational environment,a momentum-driven robot(MoRo)is proposed in the context of asteroid surface landing detection.The friction and collision characteristics are used by MoRo in the weak gravitational environment.By actively identifying the environmental parameters,the momentum wheels are planned and controlled to generate the desired driving torque so as to complete the controllable jumping tasks and stable photographing tasks after flying.Firstly,the jumping mechanism of MoRo is analyzed and the jumping method of MoRo is planned based on the characteristics of MoRo’s momentum wheel brake mechanism in the weak gravitational field.Secondly,considering the nonlinear characteristics of the three closed-loop servo system of momentum wheel driving mechanism,the Herze model and the Karnopp model are used to establish the jumping behavior dynamics model of MoRo on the surface of the asteroid.Then,the functional relationship is established between the environmental parameters and the MoRo’s motion using machine learning algorithms.The jumping distance and flying height can be controlled by the speed planning of momentum wheel based on the environmental parameters.The last,a continuous jump mission is designed.The numerical simulation validates the feasibility of the jumping planning method and the control method.
作者
王云飞
张尧
李谋
张景瑞
WANG Yun-fei;ZHANG Yao;LI Mou;ZHANG Jing-rui(School of Aerospace Engineering,Beijing Institute of Technology,Beijing 100081,China)
出处
《宇航学报》
EI
CAS
CSCD
北大核心
2021年第5期572-580,共9页
Journal of Astronautics
基金
国家自然科学基金(11972077,11672035)
国防科技重点实验室基金(6142208180103)。
关键词
小行星探测
动量驱动
跳跃行为动力学
机器学习
Asteroid exploration
Momentum-driven
Jumping behavior dynamics
Machine learning