Development of Virtual Factory Using Supervisory Control Theory

点击翻译按钮获取中文摘要

Consider a robot with six legs to walk around a mudflat for collecting environmental data. In PARI (The Port and Airport Research Institute), such a robot has been developed for the last 3 years. We have also realized its prototype using MINDSTORMS with six CPUs as shown. In the control system design, we need the two kinds of controllers, which are (1) servo controller for following a command given for each joint in each leg, and (2) gait controller to give the command to collaborate the six legs. In the paper, we are concerned with the latter case. It is well known that the gait is smoothly realized by the tripod method. However, in an emergency case to avoid some obstacle, there is a possibility that the one leg would get into a hassle with the next legs. The problem could not be solved very well by the ordinary way of generating the motion data of tripod method for several cases because there are a number of complicate cases driven by unexpected events. In our opinion, the problem should be treated as a DES (Discrete Event-driven System) control problem. From the above discussion, we are motivated to apply DES control theory to supervisory controller design for a system of robots (legs in the above), that is, a manufacturing system consisting of machines. In general, however, it is hard to calculate DES controllers by hand calculation, which means the necessity to develop a computer-aided design tool. In the paper, we present DES Analysis and Synthesis Toolbox that we developed as a collection of our original M-files using MATLAB based on the textbook by C.G.Cassandras and S.Lafortune, and show its application to a collision avoidance problem related to two machines working in a narrow space. The experimental implementation was successfully done by utilizing MINDSTORMS and Real-Time Workshop Embedded Coder.