Dynamic task allocation in an uncertain environment with heterogeneous multi-agents

Dynamic task allocation (DTA) is a key feature in collaborative robotics. It affects organizations' profits and allows agents to perform more tasks when efficiently designed. Although some work has been done on DTA, allocating tasks dynamically in an uncertain environment between heterogeneous multi-agents has rarely been investigated. The solutions proposed so far have inefficiently managed uncertainty, and none of them has utilized the semantics of heterogeneous agents' capabilities. Studies measuring the performance of these techniques on real robots are also scarce. Therefore, this paper proposes an online DTA method, which introduces new functionalities that can be applied in a real environment. In particular, an uncertain incremental cost function is developed with a distributed semantic negotiation strategy that reflects heterogeneous capabilities without needing to communicate them. The proposed method is tested in a dynamic environment and experiments on heterogeneous real/virtual robots are conducted with different numbers of agents. Different statistical and visualization tools are used to analyze the results, including bar graphs for the waiting time metrics, histograms for the waiting time frequency, scatter plots for the result distribution and variance, and critical difference diagrams for ANOVA-Tukey results. The results indicate that the proposed DTA balances allocation quality and reliability, allowing the agents to serve targets equally without neglecting certain targets at the expense of the total performance. Evidently, updating the cost incrementally allows agents to update their allocation and choose better routes to finish the task earlier. Understanding the capability also gives priority to the capable agents that complete the task faster.