Hybrid adaptive differential evolution for mobile robot localization |
| |
Authors: | Masoud Bashiri Hedayat Vatankhah Saeed Shiry Ghidary |
| |
Affiliation: | 1. Computer Engineering and Information Technology Department, Amirkabir University of Technology, 424 Hafez Ave., Tehran, Iran
|
| |
Abstract: | This paper introduces a new evolutionary optimization algorithm named hybrid adaptive differential evolution (HADE) and applies
it to the mobile robot localization problem. The behaviour of evolutionary algorithms is highly dependent on the parameter
selection. This algorithm utilizes an adaptive method to tune the mutation parameter to enhance the rate of convergence and
eliminate the need for manual tuning. A hybrid method for mutation is also introduced to give more diversity to the population.
This method which constantly switches between two mutation schemes guarantees a sufficient level of diversity to avoid local
optima. We use a well-known test set in continuous domain to evaluate HADE’s performance against the standard version of differential
evolution (DE) and a self-adaptive version of the algorithm. The results show that HADE outperforms DE and self-adaptive DE
in three of four benchmarks. Moreover, we investigate the performance of HADE in the well-known localization problem of mobile
robots. Results show that HADE is capable of estimating the robot’s pose accurately with a decreased number of individuals
needed for convergence compared with DE and particle swarm optimization methods. Comparative study exposes HADE algorithm
as a competitive method for mobile robot localization. |
| |
Keywords: | |
本文献已被 SpringerLink 等数据库收录! |