A Novel Hybrid Bacterial Foraging Optimization Algorithm Based on Reinforcement Learning

Abstract

This paper proposes a novel hybrid BFO algorithm based on reinforcement learning (QLBFO), which combines Q-learning with the improved BFO operators. In the QLBFO algorithm, under the guidance of Q-learning mechanism, each bacterium has the chance to adaptively choose appropriate one from three chemotaxis mechanisms to adjust step size. In addition, to maintain the diversity of the whole bacterial population and promote the convergence speed of the algorithm, we also improved two operators. On the one hand, we add the learning communication mechanism in the chemotaxis operator, which can make the bacterium learn from the current best one during the searching process. On the other hand, to alleviate the premature problem, a novel mechanism is adopted into the process of elimination and dispersal for each bacterium. Finally, experimental results show that the proposed QLBFO performs better than four compared algorithms.