Font Size: a A A

The Computation Of Equations Of Dynamics On Manipulator By Neural Network

Posted on:2001-08-01Degree:MasterType:Thesis
Country:ChinaCandidate:L F ShenFull Text:PDF
GTID:2168360002952414Subject:Control theory and control engineering
Abstract/Summary:PDF Full Text Request
Since 1950s the control of Robot has always been a difficulty and a focus in the kingdom of Automatic Engineering. Meanwhile, the theory of neural network has developed from the primal BP net, ADALINE to all kinds of complicated intelligent net-works such as FUZZY NEURAL NET and WAVELET NEURAL NET and so on by the constant efforts of numerous scientists. These works provide a broad view in the control of the nonlinear model, and a much better resemblance to the system. But with the increasing enlargement of the arithmetic it is urgent for us to enhance the speed of the calculating. Using many CPU to form a structure of MIMD is a common way but may bring memory conflict when dealing with many units, and the system needs huge buses. Nevertheless a much broader bandwidth is needed if it is applied. All mentioned above leads to that the application of the neural works is limited in the laboratory in most cases. On the other hands, the technology of VLSI has matured more than ever. The High Density Field Programmable Logic Devices including CPLD and FPGA are more popular in use. Although ASIC shares the same function of putting a great amounts of logic function into a small IC, CPLD and FPGA have more flexibility in that they could be used not only in short terms and also in long terms and own much higher integration degree. The paper deals with replacing the Robot kinetics with neural network by using a parallel arithmetic. Using FPGA, the network could calculate many complicated formulas in real time in a chip. This would result in the great change in the areas of robot system and its applications: We don need to exhaust our precious and limited energy to analyze and transform the complicated kinetics system, just pay our attention to behavior of the robot. Acknowledgedly, It has been long since been proved that the neural network could simulate any nonlinear systems. The paper takes the advantage of off-line study, determines the matrix of W. The input of data is the current defined by CPU, which is calculated in the electronic circuit, and the result is attained with high precision. In the next ceta the RobOt would be more widely used not only in the high-techareas such as n8tional defense, medical tredrieni, ocean developmeni, cosmosdevelopment, but also in the daily life of people. ms paPer has great significance asfor the mentioned above.
Keywords/Search Tags:Robot, Neural Network, FPGA CPLD, DPCM, ASIC, Silicon, compile
PDF Full Text Request
Related items