Fitness-Based Hierarchical Reinforcement Learning for Multi-human-robot Task Allocation in Complex Terrain Conditions.

A fitness-based hierarchical reinforcement learning method is proposed in this study for multi-human-robot task allocation in complex terrain conditions. Firstly, three fitness functions, including task fitness, distance fitness, and environmental fitness, are designed to quantify the adaptability of each subject to each task. Then, a two-layer hierarchical reinforcement learning algorithm is designed for task allocation. The values of the fitnesses are used as input for the task allocation algorithm. A series of task allocation experiments are conducted to verify the effectiveness of the proposed methods. Compared with the conventional reinforcement learning algorithm, the task allocation efficiency increases by no less than 57% by the proposed method under different numbers of subjects, and by about 65% under different numbers of tasks. [ABSTRACT FROM AUTHOR]