Q-learning-based exponential distribution optimizer with multi-strategy guidance for solving engineering design problems and robot path planning

Abstract

Exponential distribution optimizer (EDO) is an effective mathematics-based optimization technique with widespread application across numerous domains. However, it still suffers from insufficient exploitation capability, poor exploration capability, and weak adaptability of switching parameter methods. To tackle these problems, we present a Q-learning and multi-strategy guidance-enhanced EDO (QMEDO), integrating three essential innovations into the EDO framework. Firstly, the algorithm uses two local search strategies to enrich its exploitation ability and accuracy. Secondly, it employs a modified search operator with two equations to identify promising search regions effectively. Finally, a Q-learning-based switching parameter adaptive mechanism improves the adaptability of the algorithm under different conditions. The QMEDO’s performance was verified on the Institute of Electrical and Electronics Engineers (IEEE) Congress on Evolutionary Computation (CEC) 2020 and 2022 test functions. The results underscore that QMEDO demonstrates significant overall effectiveness, robustness, and convergence superiority. Then, we employed QMEDO to handle 17 engineering problems. The results indicate that QMEDO ranks first overall on these design problems and is more robust in finding the optimal solution. We further employed the QMEDO to address the mobile robot path planning problem with different initial conditions and obstacles, and the findings exhibit that our QMEDO ranks first in overall performance, has higher path planning accuracy, and better stability. The above-mentioned results reveal that in QMEDO, the collaborative effect of multiple strategies makes it reliable and applicable for solving different problems.