A Hybrid Evolution Strategy for vehicle suspension optimization

The objective of this paper is the interrelation of the diverse characteristics of two categories of optimization algorithms, stochastic and deterministic, as well as the exploitation of the advantages that each method presents simultaneously. It is well known that evolutionary methods are very efficient in avoiding local optima due to their stochastic nature, but there is no significant indication regarding to their convergence rate. On the other hand, deterministic methods cannot avoid local optima, but they converge very rapidly towards points with zero gradient. According to bibliography the evolution strategy methods have already proved their capability in avoiding local optima and focusing on much better solutions than most optimization methods. Many researchers have implemented different versions of the classical evolution strategy with good results in a wide variety of test cases. A disadvantage of such methods is the large number of iterations for the convergence of the algorithm. In the present work, the acceleration of the (1+1) Evolution Strategy with means of deterministic algorithms, such as the steepest descent method, is proposed and indication that such combination yields significantly faster convergence is produced. The test functions used for the optimization procedure are multi-variable, strongly non-linear and present many local optima. The proposed hybrid algorithm is then applied to the problem of optimizing the characteristics of a suspension system used in ground vehicles with satisfying results. Key-Words: evolution strategy, hybrid optimization, deterministic algorithms, vehicle suspension.