摘要
采用传统方法计算没有封闭解的机械臂的逆运动学运算量大、精度无法保证,对于复杂结构很难满足实时精确控制的要求;6个并行三层双隐层前馈神经网络被设计用来解决排爆机器臂的逆运动学问题,神经网络的应用受到输出误差的限制,需要减小网络输出误差;针对机械臂结构,以神经网络输出为初始值,对网络输出关节变量进行实值编码,采用分离位姿模拟退火算法对的机械臂末端位置、姿态分别进行优化;仿真结果显示,该方法有效地减小了网络输出误差,在运算结果精确性和运算速度方面满足排爆机械臂求逆运动学解的要求。
It is inadequate for tranditional methods to solve the inverse kinematics problem of manipulators which have no closed solutionsand dif- ficult to obtain real--time precision control of complex structuresbecause of the huge computation and low precision. Six parallel feedforward neural networks are designed to solve the inverse kinematics problem of a EOD manipulator. The application of neural networks is limited by output error, so the output error of networks should be minimized. Base on the structure of manipulator, the outputs are encoded with real--value and improved by simulated annealing algorithm in position and orientation. The simulation results show that the error ofend position is minized and the accuracy and calcultion speed of results meet the demand of solving inverse kinematics problem of EOD manipulator.
出处
《计算机测量与控制》
2015年第4期1269-1272,共4页
Computer Measurement &Control
基金
国家863项目(2001AA422420)
关键词
排爆机械臂
逆运动学
神经网络
模拟退火
EOD manipulator
inverse kinematics
neural network
simulated annealing