Study On Robotic Manipulator Based Neural Network And Sliding Mode Algorithms

Although the robotic manipulator is a multi-input and multi-output nonlinear system with time-varying, strong coupling and dynamic characteristics, which is difficult to achieve by conventional means, for different tasks these often require planning the space trajectory, which is particularly important of the implementation joints. Now, many scholars have proposed a variety of solutions, including hybrid control method, by which you can solve two or more fused trajectory tracking problems. By the neural network sliding mode algorithm, when the object is on sliding surface it can be out of interference, models’ uncertainties and can response fast, which is benefit, so the algorithm has been widely used and improved. Specific studies are as follows:Although traditional methods have achieved, the application has two major drawbacks:①It needs to establish precise mathematical model, and the system is difficult to deal with the uncertainty and ensure a good arm dynamic with static quality.②Controller initial output torque is too large to stand with the maximum torque. It is limited by coefficient to improve the system performance further. Although SMVSC has non-linear characteristics, the control will produce buffeting. To overcome this drawback, this article combins SMVSC with the neural network:1. Through the D-H method, on the basis of establishing needed for each robot coordinate system, through the analysis of different coordinates, the transformation matrix between the then coordinate position and posture of the relationship can be solved, then combine the joint variables through an analysis of the positive kinematics equation of the robot, and then using the method of algebra (reverse transformation method) to reverse the robot kinematics analysis, calculate the joint variables(θ1，θ2，θ3，θ4，θ5，θ6）.2. However, Sliding Mode can be used to achieve a linear and nonlinear systems robust control methods, then that object is accused of sliding surface, this time from outside interference and models uncertainties affected and fast response. Based on the characteristics of robot dynamics, in its in-depth analysis, based on reaching law proposed a new algorithm designed to improve the convergence speed of joint movement fast track the desired trajectory, reducing the system response time, weaken the system chattering increase robustness. When changing the initial position of the joint, the method can be obtained by changing the parameters of reaching law good dynamic quality through it with the traditional PID control algorithm and dual power reaching law sliding mode control algorithm are compared to verify the feasibility and reasonable sex.3. By improving the reaching law of the proposed control algorithm suppressing buffeting though achieved good results, but there are still some limitations and parameters more options randomness large neural networks to solve this problem combined with the introduction of improved reaching law control. Improved control algorithm parameters selected value while considering the introduction of periodic disturbances, by comparing the test show that the improved control algorithm even in the case of periodic disturbances can maintain robust and simple, real good.