Research On Several Swarming Behavior Based On Local Information

Swarming behavior can be found in nature in many organisms from simple bacteria to mammals. Many great progresses have been made by biologists in modeling and exploring the collective dynamics of swarm. Such behavior can result from several different mechanisms. For example, individuals may respond directly to the distribution of some biological clues as seen in some bacteria and insects, or they may respond directly to other individuals as seen in fish, birds, and herds of mammals. A challenging problem in complexity is the mathematical modeling of such swarm behaviors and its application to the control of social behavior in artificial world, such as group of mobile robots.The evolutionary process of biological swarm behavior can be looked as an algorithmic design process which is to design biological behavior to be best suit to their environment, and this process has been going on for millions of years. Therefore, general rules which govern the operation of these systems may be discovered by researching on these biological systems, and these rules have the extremely important role for developing similar engineering applications.In this dissertation, we consider models for aggregating and flocking swarms and perform scalability and stability analysis of emerging collective behavior.Firstly, we introduce the definition of swarm system and main question in this domain; the present status about the research of aggregating behavior and flocking behavior of biological swarm; reviews about modeling and control of the engineering swarm such as group of mobile robots; research about the Particle Swarm Optimization algorithms which is inspired by the interaction of individuals in the swarm.Then, we proposed a minimal circumcircle based aggregating model for scalable swarm in 2-dimensional space, and its stability properties are analyzed. The motion of each individual in the swarm is determined by two factors: (1) attraction to the center of minimal circirmcircle which is decided by the neighbors in the visual set of individual and (2) repulsion from the nearest neighbor in the repulsive range of individual. It is shown that the individuals (autonomous agents or biological creatures) will form a cohesive swarm in finite time. Moreover, explicit stable distance between the individuals and time of convergence are derived.After that, we consider flocking swarms based on the minimal circirmcircle method. We assume that the swarm is moving in an environment and propose an improved cooperative tracking model in leader-follower mode. The leader of the swarm tack the other individuals to the global object while avoiding the obstacles in the environment, and the follower track the center of the minimal circumcircle which is decided by the neighbors in the positive visual set of individual. Moreover, we analyse the scalability and stability of the flocking behavior both in free space and multi-obstacles space, and propose an evolutionary programming algorithm for swarm system.Then, we study the scalability of individual's neighborhood. By switching the control law with different attractive set of individuals, the observability between the individuals can be hold on, and scalability in both swarm layer and neighbor layer of individual can be improved. We propose a two-layer emergent topology which is fit to describe the emergent mechanism of scalable swarm. Moreover, we specify a kind of fuzzy membership function to describe the uncertainties in the motion and position informations of individuals, and analyse the dynamics of swarm behavior in uncertainty environment.After establishing the swarm model, we need to design the actual robot to make its realistic dynamics be consistent with the ideal dynamics of swarm. We consider flexible formation control of mobile robots. It is assumed that flexible formation is required to aggregate into a cohesive group and flocking to the global object. We illustrate the procedure with application to flexible formation control of mobile robots based on sliding mode control theory.In a word, it is of great significant not only for understanding the emergent mechanism of biological swarm but application of engineering swarm to study the intrinsic connection between the emergent behavior of the swarm and the interaction of the individuals in the swarm. Therefore, it is imperative to comprehensively synthesize the multi-disciplinary work achievement and develop the comprehensive method to understand the group behavior.