Evolvable Hardware Applied to Multi-objective Optimization of Controllers Structure for Robot Manipulators

The increasing complexity of the modern control systems has emphasized the idea of applying new approaches in order to solve design problems for different control engineering problems. This paper reports a performance comparison between traditional (linear PID controller) and evolvable methods (evolvable hardware controllers) applied to the problem of three-degrees-of-freedom manipulator control. An evolution procedure is used to tune the PID controllers and design the EHW controllers raised as a multiobjective optimization problem (MOP), applying the multiobjective genetic algorithm NSGA-II (MOGA NSGA-II). In this paper the objective functions are deployed considering the Integral of Square Error (ISE) criteria as performance index to be minimized. This paper shows that the evolutionary approach can discover the concept and learn the behavior of the control system. Numerical simulation results are presented to validate the controllers design and to demonstrate that the proposed evolvable controller has excellent results. Key-Words: Evolvable Hardware, Robot Manipulators, Evolvable Controller, Multi-objective Optimization.