基于神经网络的机器人碰撞检测系统研究开发  

RESEARCH AND DEVELOPMENT OF ROBOT COLLISION DETECTION SYSTEM BASED ON ARTIFICIAL NEURAL NETWORK

作　　者：王晓伟[1] 陈一民[1] 沈雄[1] 

机构地区：[1]上海大学计算机工程与科学学院

出　　处：《计算机应用与软件》2009年第7期62-65,共4页Computer Applications and Software

基　　金：上海市科技发展基金(015115042)

摘　　要：要保证机器人本身具有比较高的健壮性、安全性和实用性,机器人必须具有碰撞检测功能,给出一种采用人工神经网络进行机器人碰撞检测的原理及实现方法。在分析机器人三轴多功能位置卡有关脉冲数据的基础上,总结机器人运动连续性的规律,并建立机器人碰撞检测问题的人工神经网络模型,设计并实现了相关的机器人碰撞检测算法及控制模块,最后介绍使用多线程方法,将碰撞检测模块加入到了机器人控制系统中。Robot must have its collision detection function to ensure the higher robustness, safety and practicality of itself. In this paper, we introduced the principle and method of a kind of robot collision detection using artificial neural network. Based on the analysis of the pulse data collected from robot triple-axle multi-function control card, we got the law of continuity of robot＇ s movement, and built up the artificial neural network model for robot collision detection, and designed and realized the algorithm and control model of the robot collision detection. At last, we explained how to add the collision detection module to the robot control system using multi-thread technology.

关 键 词：碰撞检测 神经网络 BP算法 多线程 

分 类 号：TP242[自动化与计算机技术—检测技术与自动化装置] TP183[自动化与计算机技术—控制科学与工程]