Autonomous underwater surveillance networks: a task allocation framework to manage cooperation

Abstract

The design of efficient task allocation schemes is essential to manage autonomous underwater robotic surveillance networks. The network has to assign the most suited robots to the tasks which compose the mission, in spite of the unreliable and intermittent underwater communications. In this paper, we describe a market-based policy for this. It works in a completely distributed way and, through periodic auctions, sequentially allocates the robots to the tasks. There is no central auctioneer and each robot can resolve the current auction. These features increase the robustness to poor communications. The proposed scheme can manage two kinds of tasks, continuous tasks which never terminate and are always available in the task pool and sporadic ones, created upon the occurrence of some events and requiring a rapid response of the network. An Anti-Submarine Warfare scenario is considered to validate the task allocation scheme. Surveillance of areas of interest are the addressed continuous tasks, while sporadic tasks are created whenever a track produced by the tracker on-board a robot is confirmed as likely related to a target. In this case the team has to rapidly react to the event and to increase the tracking performance. Results from nontrivial Matlab simulations are reported and demonstrate the effectiveness of the allocation scheme in degraded communications conditions.