An Improved Self-Organizing Map Method for Task Assignment and Path Planning of Multirobot in Obstacle Environment

Abstract

For an obstacle workspace, an integrated multirobot task assignment and path planning algorithm is proposed by combing the improved self-organizing map(SOM) neural network and the artificial potential filed algorithm. The goal is to ensure that all targets in the environment in the presence of obstacles can be visited by a desired number of robots. The competitive layer neurons converge to the input layer neurons along the direction of resultant force in the artificial potential field to avoid obstacles effectively. The resultant force, instead of the Euclidean distance, is regarded as the winner selection criteria to find a better task executor. A novel constraint function is proposed to restrict the self-organizing properties of the SOM network. It is capable of preventing the problem of misconvergence of the neurons in the competition layer. The simulation studies with different environments and comparison results with the traditional SOM algorithm demonstrate the effectiveness of the proposed approach.