Based On Improved Particle Swarm Algorithm For Swarm Robot Characteristics

Particle swarm optimization (PSO) is a population-based stochastic optimization algorithm inspired by social behavior among animal society such as bird flocking. Due to its simple structure, fast convergence speed and easy implementation, PSO has achieved great development and has been applied to some areas. Swarm robots are a kind of robot systems, in which there are many individual robots with limited capacity and the characteristics of swarm intelligence emerge through the interaction, coordination and control of robots. Both PSO and swarm robot search are instances of multi-agent search. There exists a certain mapping between PSO and swarm robot search. For PSO, the search is abstract and virtual, while swarm robots search is specific and situated in the real world. Therefore, we can use particle swarm to model and simulate the real-world swarm robots. Related research has been reported. So we can take advantage of searching characteristics of swarm robots to improve PSO. Research in this area is quite scare, so it is necessary to do further research.Firstly, by using the features of swarm robots'asynchronous communication strategy, an asynchronous particle swarm optimization (APSO) and an asynchronous stochastic particle swarm optimization (ASPSO) are presented. In the evolution process of particles, by using asynchronous communication pattern, the information of the global best position can be asynchronously transmitted in the swarm. The simulation results show that compared with the Synchronous Particle Swarm Optimization (SPSO), improved algorithms have better local search capabilities and faster convergence rate. In addition, the effectiveness of the asynchronous pattern has been proved by theoretical analysis. Secondly, we make use of collision characteristics of swarm robots to solve the problem of premature convergence. By definition of a novel measure of population diversity function and a new concept of the particle's best flight direction, an improved attractive and repulsive PSO (MARPSO) is given. The simulation results show that MARPSO can effectively increase the diversity of population, and maintain a higher convergence speed and stronger global search ability. Then, the effectiveness of the algorithm has been proved theoretically. Finally, the inertial characteristic of robots is applied to particle swarm optimization. By reducing the search-step of the movement of particles, one-dimensional search strategy and dynamic step-size strategy are presented. The simulation results show that these strategies improved the convergence speed and local search ability of PSO.