The Study On The Multivariant Optimization Process Memorise Algorithm And Multimodal Optimization Under The Dynamic And Static Condition

Computer technology has paved the way for the intelligent optimization, Countries around the world regard the intelligent optimization as one of the development strategies. Multi-modal optimization problems with multiple optimal and suboptimal solutions constantly appeared in the production and living. However, traditional intelligent optimization algorithms cannot provide multiple optimal and suboptimal solutions in a single run. Consequently, the demand for multi-modal optimization algorithms is rapidly increasing, which makes multi-modal optimization become a research hotspot in the field of optimization. In the past ten years, although multi-modal optimization problems have been solved by some multi-modal optimization strategies, these strategies possess some limitations which make them ineffective in solving the multi-modal optimization problems under the limited computing resource or dynamic condition. To deal with this problem, this dissertation proposed a Multivariant Optimization Process Memorise Algorithm (MOA) and proved the effectiveness of MOA in solving multimodal optimization problems under the dynamic and static conditions.MOA was established by analyzing the multi-modal optimization strategies and the hotspots of multi-modal optimization. The basic idea, program flow and optimization example of MOA were detailly described to introduce MOA.To prove the effectiveness of MOA and lay the theoretical foundation of MOA for multi-modal optimization problems, This dissertation firstly estimated the complexity of the algorithm and clarified that the complexity of MOA is low; Secondly, the convergence of MOA was proved based on the Markov chain model; Thirdly, the multivariant structure, the alternate global and local optimization method and the idea of memorizing the process in MOA were introduced; Finally, the effectiveness of MOA was vivdely proved by showing the multimodal optimization processes of MOA under the static and the dynamic conditions. Experiment platform for multimodal optimization perfprmance comparison and analysis was designed and implemented, in order to, fairly compare the perfprmance of six algorithms on fifty dynamic and static multi-modal functions and the dynamic and static path planning problems.Based on forty one commonly used test function from the international IEEE Swarm Intelligence Symposium and literations, the global and multi-solution optimization performance of MOA under the static multi-modal optimization were studied, then MOA was used to solve the static multi-modal shortest path planning problems. The results show that MOA can search for multiple optimal and suboptimal solutions among multiple local traps, and it done well in the complexity, asymptotic property, accessibility in global optimum optimization tests, as well as, success rate and fitness evaluation numbers in multiple optima optimization tests, and the optimality, stability and effectiveness in path planning tests. This proved the effectiveness of MOA in solving multi-modal optimization problems under static condition.Based on eight moving peak functions and six combined dynamic multi-modal optimization functions provided in the IEEE Congress on Evolutionary Computation, the global and multi-solution optimization performance of MOA under the dynamic condition were studied, then MOA was used to solve the dynamic shortest path planning problems. The results show that MOA can search and track for multiple optimal and suboptimal solutions among multiple local traps, and it done well in the performance on locating and tracking the global optimum, off-line property and the real-time shortest path optimality. This proved the effectiveness of MOA in solving multi-modal optimization problems under dynamic condition.