System optimization using experimental evaluation of design performance

This paper presents a method of system design based upon the multi-univariate method of global optimization. Due to the fact that the method requires a relatively small number of trials to converge upon a multivariable global solution, it is practical to use experimental evaluations of the system performance as the basis of the optimization rather than a mathematical model or simulation of the system. This is advantageous since the errors inherent in a mathematical model are eliminated. Experimental results are presented for the design of a PID controller for a servo-motor control system and the optimization of robot cycle-time via link control design.