摘要
首先介绍了采摘机器人采摘臂避障问题和强化学习理论,并将强化学习方法应用到采摘臂避障问题中,构建了一平面3自由度的采摘机器人采摘臂的多Agent避碰系统。笔者研究目标是通过Agent感知采摘臂连杆与障碍物之间最小距离d和采摘臂姿态偏差角θ两方面信息,然后进行避障规划,在复杂未知环境中使其找到合适路径采摘目标。在NET平台上进行了基于强化学习的采摘臂避障系统平台的开发与仿真,对采摘臂避障系统的避障能力进行了测试分析。仿真实验表明:采摘臂避障避碰系统避障能力比较强,能够在复杂环境中采取避障措施,并准确达到指定位置。
This paper first introduces the picking arm picking robot obstacle avoidance problem and reinforcement learning theory and reinforcement learning method, which is applied to the picking arm obstacle avoidance problem, and it build a multi agent collision avoidance system with a planar 3 DOF picking picking robot arm. The goal of this research is picking arm link and the obstacle between the minimum distance d and picking arm posture deviation angle 0 of two aspects of information through agent perception, and obstacle avoidance planning, in the unknown environment to find the right path picking target. In this paper, the development and Simulation of the picking arm obstacle avoidance system based on reinforcement learning based on NET platfimn is developed and simulated, and the obstacle avoidance ability of the picking arm obstacle avoidance system is tested and analyzed. Simulation results show that the picking arm collision avoidance system and obstacle avoidance ability is relatively strong, able to take measures for obstacle avoidance in a complex environment, and accurately to reach the designated position.
出处
《农机化研究》
北大核心
2017年第11期198-202,共5页
Journal of Agricultural Mechanization Research
基金
广西高校中青年教师基础能力提升项目(KY2016LX493)
关键词
采摘臂
强化学习
姿态偏差角
Agent避碰系统
策略选择
仿真模拟
picking arm
reinforcement learning
attitude angle deviation
agent collision avoidance system
strategy selection
simulation
