Swarm Optimization Algorithms Based On Social Force Model

Swarm Intelligence (SI) systems are typically made up of a population of simple autonomous agents. SI refers to a system whereby the collective behaviors of (unsophisticated) agents interacting locally with one another and environment cause coherent functional global patterns (such as self-organization) to emerge. Interaction of individuals contains perception and reaction. Generally, perceptual behavior, reaction and self-organizing behavior are considered as the manifestation of SI. SI provides a basis with which it is possible to explore collective (or distributed) problem solving without centralized control or the provision of a global model.In nature and artificial system, there exists a lot of self-organization phenomenons. Recent years, SI inspired from the self-organization and social behavior of swarm becomes a new research area. A spectrum of swarm simulation model has been proposed to research the behaviors of individual and explain the self-organization phenomenon. On the other hand, many population based algorithms inspired by the swarm simulation models has been introduced to solve various optimization problems, such as AS (Ant System) and PSO (Particle Swarm Optimization), AFSA (Artificial Fish Swarm Search Algorithm) and so on.Althougth many swarm intelligent optimization algorithms, such as GA, PSO, have been successfully used to solve a wide range of applications, these methods also encounter a avriety of problems. Stagnation and premature convergence are the main deficiencies of the population based optimization algorithms for the reason that poor global or local search ability of algorithms. The fact indicated that there is no algorithm to meet the needs of the people and successully solved all the problems."No free lunch" theory has proven that there is no an algorithm can solve all the problems. Therefore based on different ideas from the nature, improving the existing algorithms or developing new algorithms are very necessary.Social Force Model is a kind of crowd simulation model, proposed by Helbing etal. The concept of social force is a way to estimate a pedestrian’s behavior and tendency to move in a certain direction. The social force model only uses a set of simple forces to control the movement of pedestrian. In social force model, the behaviors of each pedestrian are determined by social forces. These forces are the desired movement force, the interaction force between pedestrians and the repulsive force from the walls. The desired movement force represents the pedestrians’desire to move towards the target. The repulsive force from the walls is the force used to measure the pedestrian’s desire to avoid the wall. Social force model successfully simulates the self-organization phenomenon of crowd and pedestrian behaviors.In this paper, a novel algorithm called SFSO (Swarm Optimization algorithm based on Social Force model) is proposed to solve numerical optimization problems and an extension, MO-SFSO algorithm, for solving the multiobjective optimization problems is also introduced.The experiments in this paper indicated that SFSO algorithm has the following advantages:1) SFSO is capable of finding the multiple global optimal solutions for multimodal function in parallel searching way;2) The search mechanism of SFSO algorithm keeps the diversity of the population in a high level, and this can prevent the algorithm trapped into a local optimum;3) The mechanism of SFSO guarantes the algorithm has strong robustness in dealing with different types of optimization problems.The SFSO algorithm can overcome the deficiencies which the PSO algorithm has and be able to make a good balance between the global and local search. This is because:1) The mechanism of SFSO is that pedestrian’s movement is driven by social force and pedestrian’s movement to the target is also the evolution process of solving optimization problems. In this process, two classes of pedestrians:free individual and nonfree individual, will be divided based on the distances of their current positions to the targets. And different approaches are utilized to increase the diversity of population by free individual and nonfree individual in exploration phase. Random search behaviors of free individuals enhance the global search ability of SFSO algorithm. As a repulsion factor, repulsion force between nonfree individuals provides a way to alleviate the premature convergence problem.2)In exploitation phase, the cooperation of pedestrians accelerates the search process and improves the accuracy of solutions. So, different behaviors of pedestrians in SFSO establish proper balance between global and local search.The search mechanism of MO-SFSO is same to SFSO and the movements of pedestrians are also driven by social force. The comparision results of MO-SFSO to NSGA-Ⅱ suggest that MO-SFSO performs well in terms of both convergence metric and diversity metric. This superiority can be attributed to the fact that MO-SFSO has excellent global search ability and local exploitation ability.