Improved metaheuristics using guided local search and its application to control of unstable system with delay

Abstract

Current research interest in Optimization over last 20 years newlinefocusses on using the merits of metaheuristic algorithms mainly their global newlinesearch ability and nature of being derivative free. They overcome the known newlinedrawbacks of derivative methods such as computational difficulty and newlinedependence on gradient information and limitations on porting legacy code newlineinto digital computers. There exist several methods of improving the search newlineability of the Metaheuristics. In this research work, the local search and global newlinemethods of optimization are combined to yield more effective hybrid newlinemetaheuristic algorithms with improved local search capability. The hybrid newlinemetaheuristic algorithms are evaluated for performance using applications solved newlineusing multi-objective optimization, unconstrained and constrained optimization. newlineA multi-objective hybrid metaheuristic is proposed using a re-inserter based Fast newlineNon dominated ordering genetic algorithm (R-NSGA-II). It initializes the newlinepopulation and processes that population using unique fitness selection, newlinecrossover and mutation mechanisms. The population used by this mechanism newlineis deployed with local search mechanism with a recent gradient free method newlinecalled Direct Multi Search (DMS). This algorithm is applied to standard test newlineproblems with known pareto front and results are compared with the results of newlineNSGA-II. Improved performance of NSGA along with proven measures for newlineimproving convergence and diversity has been presented and the results are newlinediscussed through comparison with standard test problems. The results show newlinethat the R-NSGAII overall yields higher sensitivity and robustness with newlinemaximal values of hypervolume scores and minimal IGD index for majority newlineof the problems. newline newline