DESIGN USING GENETIC ALGORITHMS OF HIERARCHICAL HYBRID FUZZY-PID CONTROLLERS OF 2-LINK ROBOTIC ARMS

Fuzzy-logic controllers (FLCs) have been shown to be very promising in controlling ill-modeled and complicated systems. Moreover, they offer an alternative to more traditional robot control schemes, and this al ternative can be more readily integrated with the artificial intellige nce required for task planning and decision making, both crucial to ro botics. However, the traditional methods of designing FLCs are based o n expert heuristic knowledge and trial and error, and are often tediou s and unyielding. In this article, we develop a computer-implemented p rocedure for designing a hierarchical hybrid fuzzy-PID (HHFPID) contro ller for the position and trajectory control of a two-link robotic arm . This procedure combines genetic algorithms (GAs), expert knowledge, and fuzzy learning from examples. We will discuss the computational is sues of our approach, and the design of fitness functions and encoding schemes required by the genetic algorithms. Based on extensive simula tion studies, we conclude that the GA-designed controller has a satisf actory and sometimes superior performance. (C) 1997 John Wiley & Sons, Inc.