A Neural Network Based On Mind Evolution Algorithm Applying To Inverse Kinematics Problem Of Robotic Manipulator

As a high-tech automechanism robot could substitute for human to complete a lot of complex, heavy and dangerous tasks such as spray paint, pick-place, soldering, assembly, medical and so on. It improves the productive efficiency and quality of product and ameliorates the work conditions.Robotic control problems include trajectory planning problem (TPP, in short), inverse kinematics problem (IKP in short) and inverse dynamics problem (IDP). The IKP is the vital part of robotics and the base of robotic dynamics and control, which is directly related to the kinematics analysis, off-line programming and so on. From the viewpoint of robotics control IKP is a main a question for discussion, somany people pay more attentions to it.At first, the robot kinematics equations are analyzed and the SCARA robotic kinematics models are established. Secondly, the back propagation (BP, in short) algorithm and neural networks based on genetic algorithm (GA, in short) are analyzed and applied to solve IKP. The theoretic pith, main arithmetic, its application and development of mind evolutionary algorithm (MEA, in short) are mainly introduced. Especially a mind-evolution method with binary code, similartaxis and dissimilation is introduced.Using the approaching ability of mapping of neural networks to non-linear function, though training lots of data implement robotic the non-linear mapping from the working variable coordinates to the joints variable coordinates in order to solve IKP. If the outputs of neural networks are changed but the activation function cannot be changed, the only way is to change the weights of the inputs. So the processing of training nets is the processing of updating the weights. MEA is presented to update the weights of the neural networks. The motivation of this approach is to overcome the shortcomings of tradition BP and GA, such as the low precision of the solutions, the slow search speed and easyconvergence to the local minimum points. In order to compare these algorithms simulations are provided. The theoretical analysis and simulations show that the proposed method applying to robotic IKP is feasible and it accelerates the convergence speed and heightens the precision of the solution. Finally, by using the results of IKP the method is applicable for robot to complete the pick-place task.