Efficient multi-AUV cooperation using semantic knowledge representation for underwater archaeology missions

Advances in the fields of communication technology and software, electrical and mechanical engineering enable the replacement of a single robot by cooperative robotic team in highly demanding applications, such as search and rescue. A robotic team could perform better than a single robot, if certain challenges, such as action planning, coordination, and decision making, are successfully tackled. One key factor for the successful performance of a robotic team is the multi-robot task allocation. Specifically, the challenge is to define which robot executes which task, considering an efficient solution for the successful completion of the complex mission. The task allocation could be even more challenging when real-world communication constraints and uncertainties are presented, such as limited bandwidth, high latency and high packet loss. In the current study, we attempt to resolve the issue of a cooperative robotic team under communication constraints. To reach this goal, the use of a distributed world model for multi-robot task allocation is proposed. This ontology based distributed world model is capable of successfully handling to a great extent the aforementioned communications limitations, thus allowing successful mission execution even under harsh communication conditions. An efficient centralised task allocation mechanism, using k-means clustering, is described, and its performance is compared to a greedy centralised task allocation method. Experimental simulation results indicate that the efficient method performs better on average than the greedy one, without extra time requirements.