基于经验学习和触觉反馈的抓取力矩控制方法研究  

Grasp torque control based on experiential learning and haptic feedback

作　　者：安慰宁 张文昌[1] 陈卓[1] 苏卫华[1] 

机构地区：[1]军事医学科学院卫生装备研究所,天津300161

出　　处：《医疗卫生装备》2017年第2期122-126,130,共6页Chinese Medical Equipment Journal

基　　金：天津科技支撑计划重点研发计划(16YFZCSF00590);军事医学科学院创新基金(2015CXJJ020)

摘　　要：目的 :为了使机械手可以完成灵巧操作,提出一种基于经验学习和触觉反馈的抓取力矩自动控制方法。方法:建立经验库,在抓取任务时通过触觉反馈识别目标物,查询目标物是否在经验库中,若在,输出力矩;若不在,通过迭代学习策略计算出来。结果:对于以往经验任务的抓取,机械手可以很快得到经验力矩输出;对于新的目标物抓取,通过迭代学习,机械手可以计算出最优力矩输出,并完成稳定抓取。结论:通过增量式经验学习,机械手经验越来越多,在遇到抓取任务时可以快速计算出力矩输出。Objective To propose a grasp torque control based on experimental learning and haptic feedback to facilitate the manipulator in dexterous manipulation. Methods An experience database was built firstly, and then the object was recognized by tactile feedback in the grasp task. If the object had been experitentially grasped, the torque was output based on the database. In case a new object was grasped, the optimal output torque was calculated by iterative learning. Results The experiment showed that the robot hand could find the experiential output torque quickly when encountering the object in database and calculate the torque by iterative learning to achieve grasp task. Conclusion The experiential database grows up when the robot hand learns more and more experience. It can fast output torque like human in the grasp task.

关 键 词：经验学习 触觉反馈 抓取力矩 预抓取 机械手 

分 类 号：TP241[自动化与计算机技术—检测技术与自动化装置]