Particle Swarm Optimization Based on Hybrid Kalman Filter and Particle Filter
- PDF / 547,197 Bytes
- 8 Pages / 612.284 x 810.709 pts Page_size
- 73 Downloads / 227 Views
Particle Swarm Optimization Based on Hybrid Kalman Filter and Particle Filter PENG Pai 1 ( ),
CHEN Cong 2 ( ),
YANG Yongsheng 1∗ ()
(1. School of Aeronautics and Astronautics, Shanghai Jiao Tong University, Shanghai 200240, China; 2. Key Laboratory of Integrated Technology of Avionics System, Shanghai 201103, China)
© Shanghai Jiao Tong University and Springer-Verlag GmbH Germany, part of Springer Nature 2020 Abstract: The combination of particle swarm and filters is a hot topic in the research of particle swarm optimization (PSO). The Kalman filter based PSO (K-PSO) algorithm is efficient, but it is prone to premature convergence. In this paper, a particle filter based PSO (P-PSO) algorithm is proposed, which is fine search with fewer premature problems. Unfortunately, the P-PSO algorithm is of higher computational complexity. In order to avoid the premature problem and reduce the computational burden, a hybrid Kalman filter and particle filter based particle swarm optimization (HKP-PSO) algorithm is proposed. The HKP-PSO algorithm combines the fast convergence feature of K-PSO and the consistent convergence performance of P-PSO to avoid premature convergence as well as high computational complexity. The simulation results demonstrate that the proposed HKP-PSO algorithm can achieve better optimal solution than other six PSO related algorithms. Key words: Kalman filter, particle filter, particle swarm optimization (PSO), intelligent algorithm CLC number: TP 183 Document code: A
0 Introduction Intelligent algorithms have been used effectively to reduce the complexity of real-world optimization problems[1] , such as hard numerical functions, which are usually exponential. Particle swarm optimization (PSO) is a kind of swarm intelligence algorithm based on stochastic population. It was first introduced by Kennedy and Eberhart[2]. They discovered the mechanism of PSO through the intelligent behavior of unintelligent creatures[3]. In the original PSO[2] , a group of particles flies in the search space. The initial coordinates of particles correspond to randomly generated solutions, where each particle learns from the best record of all particles. As the number of iterations increases, particles tend to approach the position of the particles with the best fitness. The process of iteration can be simply expressed by velocity update function and position update function. PSO is inspired by the social and biological behavior of bird flocks searching for food sources. Mendes et al.[4-6] revealed how each particle learns from its neighbors and proposed the fully-informed PSO in which each particle learns from its neighbors rather than simReceived date: 2019-05-20 Foundation item: the Aviation Science Foundation of China (No. 20165557005) ∗E-mail: [email protected]
ply from the best performer. Shi and Eberhart[7-8] introduced inertia weight into the original PSO to achieve the global optimum within a reasonable number of iterations easier. Clerc and Kennedy[9] proposed random coefficients to PSO to guarantee convergence. When the
Data Loading...
