Multi-strategy-based artificial bee colony algorithm for AUV path planning with angle constraints

Abstract

Listen

Translate article

This paper focuses on a path planning problem of an autonomous underwater vehicle (AUV) traversing target points at desired angles in an obstacle environment with eddy currents. A tangent-spatial partition method for path planning model construction is developed, which guarantees the direction of arrival in addition to extending the scope of waypoint search. An improved artificial bee colony algorithm integrating with multiple evolutionary strategies (ABC-MES) is proposed. More specifically, a population initialization method that follows the best point set principle is presented to increase population diversity and accelerate convergence. A multi-strategy evolutionary approach is developed in the employed bees phase to enhance population quality and establish a decision support database for subsequent upper confidence bound (UCB) learning. The UCB algorithm is then applied in the onlooker bee stage for assessment and screening of the optimal evolutionary strategy according to the accumulation of prior knowledge and the exploration of new knowledge. Finally, the T-distribution and reversal learning are explored in the scout bee stage to update old nectar sources to prevent the algorithm from falling into the local optimal prematurely. The strong ability of the proposed ABC-MES algorithm to jump out of the local optima is ensured by comparing it with 9 existing algorithms in terms of accuracy and stability with the help of 28 test functions. Simulation results reveal that the path generated by the proposed ABC-MES algorithm has lower overall cost accounting for time efficiency, navigation distance and energy consumption.