Multi-type task allocation for multiple heterogeneous unmanned surface vehicles (USVs) based on the self-organizing map

Abstract

The allocation of multi-type tasks for heterogeneous unmanned surface vehicles (USVs) is the main focus of this paper. In order to consider the heterogeneity of USVs in the process of task allocation, a fitness function is established to model the differences between USVs in terms of mission execution capabilities and energy reserves. An improved self-organizing map (SOM) is proposed based upon the fitness function to assign multi-type tasks to the specific USVs. Meanwhile, for some special tasks that need to be performed in sequence, such as sea sweeping tasks, the tasks treatment list (TTL) method is introduced to take into account the precedence constraints during the task allocation. Furthermore, a variable task value model is proposed, which includes the fixed part and the time-varying part, to conform to the practical significance of the rewards obtained from the task execution. And the improved transition probability function from the ant colony optimization (ACO) algorithm is integrated into the task allocation method to maximize the overall rewards of the task execution. Three numerical simulations are carried out herein. The first two are performed to verify the effectiveness of the proposed task allocation method in a large-scale task assignment scenario while the third simulation presents the numerical results of comparing the proposed method with other benchmark methods, which are applied in an authentic maritime application. Through numerical simulations, multi-type tasks can be properly assigned to specific USVs with corresponding execution capabilities considering both distance costs and energy reserves. Under the effect of the proposed method, tasks with precedence constraints can be effectively assigned to the USVs for sequential execution and the maximization of overall rewards can also be optimized. The method has a good efficiency in a complex mission scenario and can be adjusted manually by choosing suitable weight parameters of the algorithm to meet the needs of practical applications.