An Evolvable Driver for a Non-Linear Damped Pendulum

摘要

In this work, an evolvable hardware technique has been applied to a real control environment composed by a damped pendulum. A combinatorial 4 bit circuit input/output driver design by genetic algorithms will be implemented in an Altera Cyclone II FPGA, aiming to prove the efficiency of the evolvable design. A proportional driver tuned by a genetic algorithm (Kp-AG) was also implemented and subjected to the same tests to compare their performance against the Evolvable Hardware design. Experiments use a square reference signal to force overshoot errors, and, a sinusoidal signal to sample the response for both drivers. The results obtained by the experiments show that the evolutionary approach to design drivers can be competitive and improve considerably the control in this particular case of nonlinear control using low resolution feedback information.