Strength Learning Particle Swarm Optimization for Multiobjective Multirobot Task Scheduling

Yongchun Fang,Xiao-Fang Liu,Zhi-hui Zhan,Jinchao Zhang

DOI: https://doi.org/10.1109/TSMC.2023.3239953

2023-07-01

IEEE Transactions on Systems, Man, and Cybernetics: Systems

Abstract:Cooperative heterogeneous multirobot systems have attracted increasing attention in recent years. They use multiple heterogeneous robots to execute complex tasks in a coordinated way. The allocation of heterogeneous robots to cooperative tasks is a significant and challenging optimization problem. However, little work has gone into scheduling large-scale cooperative tasks with precedence constraints and multiple conflicting optimization objectives. Existing methods are insufficient to address the issue. We propose a multiobjective model and develop strength learning particle swarm optimization (SLPSO) to optimize multiple objectives. In this article, the problem is converted into a two-step problem of task permutation construction and robot subset selection. In order to coordinate with the time-extended property of the problem, SLPSO utilizes a hybrid encode scheme: an element-based representation for task permutations and a binary representation for robot coalitions. A strength learning strategy with heuristic information guides particles to enhance their best-performing objectives for improving swarm convergence. In addition, an estimation-based local search is developed to improve spare solutions for enhancing swarm diversity, which determines the search direction by estimating fitness improvements. Experimental results on thirty problem instances are elaborated to demonstrate that the proposed SLPSO significantly outperforms the state-of-the-art algorithms in terms of inverted generational distance and hypervolume metrics. The proposed SLPSO can obtain a set of high-quality and diversified solutions.

Engineering,Computer Science