基于神经网络干扰观测器和滑模控制的机械臂轨迹跟踪研究

打开文本图片集

Abstract： In order to solve the influence of uncertainty factors such as external disturbance，dynamic errorand modeling uncertain parameter errors on the trajectory tracking performance of the robot， a super-twisting sliding mode trajectory tracking control method based on a neural network disturbance observer is proposed. Inview of the uncertainty factors in the robot system，a neural network disturbance observer with a double hidden layer structure is designed. In order to eliminate the chattering，a sliding surface function and a super-twisting sliding mode control law are constructed to achieve the stable output of the torque of the robot system and the asymptotic stability of the system is proved by the Lyapunov theory.The proposed neural network disturbance observer is combined with the super-twisting sliding mode controller （STSMC） designed in the paper， sliding mode controler （SMC）， non-singular terminal sliding mode controller （NTSMC） and fast non-singular terminal sliding mode controller （FNTSMC）. And trajectory tracking experiments are carried out by above methods. The experimental results show that compared with the SMC，NTSMC and FNTSMC methods，the proposed method not only reduces the average maximum tracking errors of the six joints of the robot by about 32.1% （202 27.0% and 4.3% respectively，but also significantly eliminates the joint chattering phenomenon， which can achieve accurate and stable trajectory tracking control of the robot.

Keywords： neural network; sliding mode control; robotic arm; trajectory tracking

0 引言

机械臂是一个非线性和耦合性的复杂系统，在实际的运行过程中经常存在干扰和未知误差，大大增加了精确控制的难度Ⅲ，为解决该问题，众多学者提出PID控制[2]、模糊控制[3]、滑模控制[4]、自适应控制等控制方法[5]。（剩余9901字）