Multi-rule Artificial Physics Optimization Algorithm

Heuristic optimization algorithms provide an effective solution for solving optimization problems, however, most of the current algorithms only follow a law of motion in an iterative process, so that all individuals follow the same motion characteristics and can’t reflect balance mining and exploration capabilities effectively. This paper study as the Artificial Physics Optimization algorithm, proposed a Multi-rule Artificial Physics Optimization algorithm.First, this paper constructed a force rule set and put forward two kinds of the equiprobable random selection strategies that which rules to be choosed, including random selection strategy based on population and random selection strategy based on the individuals. The selection strategy that based on population will distribute a certain number of individuals equally initially. Then allocate more individuals to the force rule that lead more individuls to the better. And the selection strategy based on the equal probability that each individual select any force rules, in the next generation, the selection probability of the individuls to force rules will be increase or decrease in accordance with the fitness being better or worse.In order to verify the validity of these two algorithms, compared with the standard APO and an extension APO algorithm.And the experimental results also show the performance of these two kinds of algorithm.Later, according to the principle of physics that the different states follow the different motion, this paper proposes a multi-rule Artificial Physics Algorithm based on population diversity. According to the need of the algorithm, this algorithm use the temperature controlling environment, change the physical condition of the individual, and make the individual moving with the corresponding reaction rules, The species diversity judies the convergence of the algorithm and making the algorithm stable in the end. In order to verify the effectiveness of the proposed algorithm, compared with the equal probability random selection strategy of APO algorithm、MEABC、PSO and TVAC(the improved PSO algorithm), the result shows that the effectiveness of the APO algorithm based on population diversity.Finally, the Multi-rule Artificial Physics Optimization algorithm will be applied to optimal design of the PID controller. Improved the disadvantages of manual control parameters of the PID controller, the more close to the ideal type of step curve.