Learning Automata-Based Multi-target Search Strategy Using Swarm Robotics

Abstract

Swarm robotics is widely studied in multi-target search problem because of its low cost and adaptability in dangerous environments. But current multi-target search strategies have the problem of searching the same area repeatedly and are difficult to search the undetected area effectively. This paper proposes a learning automata-based multi-target search strategy (LAS). The strategy divides the search space into multiple cells and initializes each cell with an equal search probability. The probability distribution of cells is learned and updated by a learning automaton and employed to assign robots to search cells. If a robot detects the presence of a target in an assigned cell, it uses the simulated annealing algorithm to search the exact location of the target. The experimental results demonstrate that the proposed strategy significantly improves the search efficiency compared with the state-of-the-art methods.