Robot Manipulator Identification based on a Customized Neural Network & Close Form Differential Equation

In this paper we try to explore the potential of using dynamic neural network approaches to identify the internal behavior of structure-unknown non-linear time variant dynamic systems. We suppose that simple structure or a close form of plant differential equations is available. These types of identifications typically encountered in the field of applied mechanics where the system behavior is fully dynamic and describe by differential equations. In this work, we use the relevant characteristics of recurrent neural networks in serial-parallel topology fashion. Some additional derivation vector of plant input/output driven from the plant differential equation is used as additional input to the network. The network consists of two isolated sections combined with a single neuron. It is shown that, using this topology, the error rate of modeling has been decreased and therefore the identifier performance and resolution increase to the levels reported in others works. We use industrial robot manipulator for the case study in this work. The manipulator is simulated with professional simulation software (consist of solidwork, visual nastran 4D and matlab/simulink).