Precision Measurement Institute made progress in the study of intelligent optimization algorithms

[ Instrument R&D of Instrument Network ] There are many problems in geodetic research. Particle swarm optimization is a search-based stochastic optimization method based on population, which is used to solve the optimal solution of single/multi-objective problems. Due to the advantages of strong robustness, fast convergence speed, and few adjustable parameters, particle swarm optimization has attracted the attention of researchers in different fields in recent years. However, the existing particle swarm optimization algorithm has the defects of insufficient population diversity, premature convergence and easy to fall into local optimality. When the optimization problem has a large number of local optimal values ​​or the dimension is high and inseparable, the solution effect is poor.

Recently, the application group of geodetic new technology application of the Institute of Precision Measurement Science and Technology Innovation, Chinese Academy of Sciences has made progress in the research of particle swarm optimization algorithms in swarm intelligence optimization. The study first proposed that the overall population be divided into two heterogeneous subgroups (comprehensive learning strategy subgroup and dynamic multi-group subgroup), in which comprehensive learning strategy subgroup is mainly responsible for development, and dynamic multi-group subgroup is mainly responsible for exploration; second, Classify the search capabilities of multiple dynamic subgroups, and construct a new adaptive nonlinear decreasing decreasing inertia weight based on the classification results. Finally, two mutation operators (non-uniform mutation and Gaussian mutation) are introduced to improve the locality of the algorithm Seeking the best ability.

The research team evaluated the performance of the proposed HCLDMS-PSO algorithm through two international standard optimization problem test sets (CEC2005 and CEC2017) and an actual wireless sensor network coverage optimization application problem, and compared it with the 8 international advanced The particle swarm algorithm variant is compared with other 11 species intelligent optimization algorithms. The results show that the new algorithm effectively improves the convergence speed, optimization accuracy and reliability on most optimization problems. The intelligent optimization algorithm is expected to be applied in the fields of mobile 5G positioning, intelligent driving, and image matching positioning.

Related achievements were published in Information Science under the title of Heterogeneous comprehensive learning and dynamic multi-swarm particle swarm optimizer with two mutation operators. The first author of the paper is PhD student Wang Shengliang, and the corresponding author is researcher Liu Genyou. The research work was jointly funded by the National Key R&D Program and the National Natural Science Foundation of China.