Abstract
This investigation uses evolutionary algorithms (1, 2) as part of a methodology for suspension control system design. It investigates both active and semi-active systems. Active suspensions allow power to be applied to the suspension while semi-active systems allow only changes in suspension characteristics without power input. These control algorithms are investigated using non-linear modelling techniques and are not constrained to linear or root-mean-square (RMS) optimization criteria. Jerk (rate-of-change of acceleration) is used to determine ride smoothness and a non-linear function is used to determine how well the suspension stays within the rattlespace (the limits of travel of the suspension). The systems are also tested over a large number of random road contours which mimic the characteristics of actual roads. The comparison of the various control programs indicates how well they combine the objective of ride smoothness with the objective of staying safely within the physical limits of the suspension travel.
Each of the suspension control algorithms presented below has real-valued parameters that are adjusted by evolutionary algorithms (EAs) to achieve an optimal or a suboptimal response to random road contours. A computer program, developed as part of a PhD research project, is the test bed for the EAs. Preliminary results of the research are presented in a previous paper (3) in which various algorithms were tested. None of the suspension algorithms of that paper performed better than the passive suspension. In the experiments below, the passive suspension the simple skyhook and the on-off skyhook algorithms are used as benchmarks, and some of the control algorithms presented below show significant improvement over all these benchmarks.
Keywords:Control, suspension, jerk, multiobjective opimisation, evolutionary algorithms.