ANN Based Formation Control of Intelligent Multi-Robot Systems With SoPC Implementation

• Main Authors: Tai-Chun Chiu, 邱太鈞
• Other Authors: Sendren Sheng-Dong Xu
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/c44c2p

碩士 === 國立臺灣科技大學 === 自動化及控制研究所 === 107 === This thesis presents the formation control method of the multi-robot system, constructed by three four-wheeled omnidirectional mobile robots, and its implementation by using the system on programmable chip (SoPC). First, we analyze the kinematics and dynamics of a single robot, and then design the controller. Models of Radial Basis Function Neural Network (RBFNN) and Multilayer Perceptron (MLP) are also incorporated with the controller to make it become an adaptive controller able to adjust its control parameters with controlling behavior. Several heuristic algorithms like Genetic Algorithm (GA), Firefly Algorithm (FA), and their variation are introduced to achieve the optimization of parameters of RBFNN and MLP models. We also compare the search performance of different algorithms. After finishing the design of a single robot, we then design the formation control algorithm of the multi-robot system by using transformation of coordinates and geometry. Such design can enable the robots to track their desired trajectory while maintaining their formation shape at the same time. The development of robot’s hardware is based on DE1-SoC development board produced by Altera Corporation to fulfill the design of the system with SoPC technology. TCP/IP protocol and Socket technique are also used to construct connection between robots. Finally, simulation results by using MATLAB and physical experimental results show the correctness and advantages by using proposed methods.