Social Community Based Swarm Intelligence Optimization Algorithms With Applications In Logistics

With the rapid development of the Chinese economy, logistics as the main pillar industry of domestic economy, is now facing informatization and intellectualization transformation. In order to deal with challenges of rational utilization of logistics resources and achieve high efficiency, satety and low carbon emissions, more requirement are proposed. Under such situation, how to solve new complex logistics problems which are large-scale, dynamical and multi-objective become an important problem. Actually, conventional and classical methods could not meet the new requirements for applications. Swarm intelligence optimization(SIO) provides a feasible and efficient technology to solve complex operational problems in logistics because of its high computing performance and simple modal. However, there are still some existing problems in SIO algorithms, such as being easily dropped into local extreme areas and premature convergence, which limit their applications in logistics industry. Thus, if the shortcoming can be overcomed, SIO will release more power and be applicable to more problems. This dissertation aims at analyzing and constructing SIO algorithms from three important aspects: population topology, neighborhood structure and individual behavior, so as to improve the optimization ability of SIO in complex problems and enhance the application research in logistics. The main contents and fruits are listed as follows:(1) Considering the neighborhood structure, this dissertation proposed an adaptive particle swarm optimization based on clustering(APSO-C). By a K-means clustering operation, it divided the swarm dynamically to variable clusters and adopted a ring topology for information sharing among these clusters. Then, an adaptation mechanism was proposed to adjust the inertia weight of all individuals based on the evaluation of the states of clusters and the swarm, thereby giving the individual suitable search power. The experimental results of benchmark showed that APSO-C had a better performance in the terms of convergence speed, solution accuracy and algorithm reliability in comparison with several other algorithms.(2) Based on the analyses of social network of population, a new algorithm called particle swarm optimization with dynamic topology based on social network evolution(PSODT-SNE) is provided. It introduced social network evolution behavior to the population to build a dynamic interaction method among individuals and constructed a dynamic population topology structure with social characteristics. Through this way, it adjusted the information diffusion routes among individuals to enhance the interactive activities of individuals and improve their search ability. Numerical experiments results demonstrated that the proposed algorithm had an excellent performance on search ability and adaptability.(3) This dissertation proposed a new swarm intelligence algorithm named social network-based swarm optimization algorithm(SNSO) based on the comprehensive influence of population topology, neighborhood structure and individual learning behavior. In SNSO, three different strategies of social network evolution rule, extended neighborhood structure and comprehensive learning behavior, were provided to improve the search ability. The effect of these aspects on the search ability were also analyzed via a series of numerical experiments. Through benchmark tests, SNSO showed a better performance compared with some peer algorithms. Importantly, it provided a feasible method to improve or develop SIO algorithms.(4) We applied the proposed algorithms to solve the container multimodal transportation problem. This work proposed a heuristic strategy based on prorate distribution of flow and local traffic adjustment to build a mapping modal between individual representation and multimodal transportation schedule. Then, PSODT-SNE and SNSO were applied to solve a series of cases with different scales. The results showed that the proposed algorithms had a superior performance on the terms of convergence speed and solution accuracy in comparison with the selected algorithms and the decoding strategy was aslo available and efficient for other peer algorithms to solve the multimodal transportation planning problem.(5) Finally, container vessel stowage planning problem was researched based on SIO algorithms. In this work, under a decoding strategy based on positions and a loading strategy, we decode an individual to a feasible stowage plan. Then the SNSO was applied to solve the problem. Through different scale cases tests, the results demonstrated that SNSO showed better performance in the terms of convergence speed and solution accuracy than seven selected algorithms and the proposed algorithm was an effective method for solving operational complex problems.