Co-evolutionary Multi-Colony Ant Colony Optimization Based on Adaptive Guidance Mechanism and Its Application

Ant colony optimization has insufficient convergence and tends to fall into the local optima when solving the traveling salesman problem. This paper proposes a co-evolutionary multi-colony ant colony optimization (MCGACO) to overcome this deficiency and applies it to the Robot Path Planning. First, a dynamic grouping cooperation algorithm, combined with Ant Colony System and Max-Min Ant System, is introduced to form a heterogeneous multi-population structure. Each population co-evolves and complements each other to improve the overall optimization performance. Second, an adaptive guidance mechanism is proposed to accelerate convergence speed. The mechanism includes two parts: One is a dynamic evaluation network, which is used to evaluate and divide all solutions by the evaluation function. The other is a positive-negative incentive strategy, which can enhance the guiding role of solutions with higher evaluation value. Besides, to jump out of the local optima, an inter-specific co-evolution mechanism based on the game model is proposed. By dynamically determining the optimal communication combination, the diversity among populations can be well balanced. Finally, the experimental results demonstrate that MCGACO outperforms in terms of solution accuracy and convergence. Meanwhile, the proposed algorithm is also feasible for application in Robot Path Planning.