Particle Swarm Optimization Algorithm To Improve Research And Its Application In The Inverted Pendulum Control System

The particle swarm optimization (PSO) was put forward by Kennedy and Eberhart, which is a kind of intelligent swarm optimization algorithm. It’s idea derived form studying birds social modeling and imitating birds behavior. The algorithm is simple, without gradient information, easy to implement, and can adjust the global and local search capabilities flexibly, it has become one of the research hotspot in random search. PSO has a convergence speed in the front, but later the problem of poor local search ability and premature convergence will emerge.In order to overcome the problem on poor local search ability and premature convergence in PSO algorithm, the dynamic random search technique (DRST) is introduced to optimize the currently best position of the swarm, and hence, the local convergence is greatly enhanced. Reinitialization with a good-point set manner is employed for the swarm falling into premature convergence to go out of the local optimum. Linear search in the negative gradient direction is also applied to accelerate the optimization. In the end, the numerical experience results show that our improved algorithm have better convergence rate, great ability of preventing premature. convergence and superior performance than the existing ones.Evolutionary algorithm always has the limitations to deal with the constraints conditions in solving constrained optimization problems. A new PSO algorithm is proposed to settle the problem. The fitness function for feasible and infeasible solution is redesigned separately. It is easy to manipulate constraint conditions by the new function. In addition, a dynamic semi-feasible region is used to increase the possibility of infeasible solution’s survive and elevate the search ability around the boundary. The numerical experience results show the effectiveness of the proposed algorithm.The controller parameters of double inverted pendulum are offline optimized by the improved PSO algorithm. The simulation show that the system can get better performance on stability, overshoot and response speed for system response.