By Super User
A framework for CASW network to integrate cooperative area exploration, track prosecution and communications optimization tasks. Ferri, Gabriele. CMRE-FR-2017-006. November 2018.
This report investigates how to improve the cooperation of the robotics nodes composing the CMRE ASW multistatic network. We propose a framework to manage the allocation of the robotic nodes to the tasks of an ASW mission. Examples of these tasks are area exploration/surveillance, track prosecution and communications optimisation. This problem is known as the Multi Robot Task Allocation (MRTA) in the robotics community. Its effective solution in the addressed ASW scenario is the key driver to exploit the benefits of cooperation. The underwater scenario (and in particular the littoral one) raises big challenges from the task allocation perspective. The limitations of underwater communications (low throughput, low range and low reliability) makes the exchange of messages between the robots and the C2 centre intermittent and sometimes impossible. Centralised control of the network, which could allow optimal task allocation, is not possible. We propose a market-based framework as a viable solution to task allocation for the CMRE network. The framework is general and considers two categories of tasks, continuous and sporadic, which are sufficient to model the different AUV operations of an ASW mission. Continuous tasks (e.g. area surveillance) do not terminate and remain always available in the team task pool. Sporadic tasks, on the other hand, are created upon the occurrence of some events (e.g. track prosecution) and require a rapid response of the network. The task allocation scheme works in a completely distributed way and two parallel auctioning systems are run in parallel to manage the two task categories. All the robots are sequentially allocated to tasks through periodic auctions. The algorithm uses negotiations among neighbouring nodes and requires only local communications. There is no central auctioneer, and each robot is able to resolve the current auction exploiting its local knowledge and the information received by the other nodes. Particular attention has been paid to make the auction scheme robust to intermittent communications. Each robot waits for messages from collaborators for several consecutive communications time frames before making a decision. The proposed task allocation policy also allows task reallocations to handle the evolution of the tactical scene. In this work we describe the general framework. We demonstrate it considering the area surveillance as an example of a continuous task and track prosecution for sporadic tasks. Results from non-trivial Matlab simulations are reported. The ARTEMIS (Adiabatic Reverberation and Target Echo Mode Incoherent Sum) model is used to create the sensor model adopted in the simulations. The simulations were set in a realistic LCAS16 scenario. Results show the effectiveness of the framework to find the task allocation with different task categories and its robustness to poor communications. We conclude with preliminary results from ASW-ODC17 trial, in which the first building block of the framework was tested at sea on CMRE OEX AUVs. In this trial the CMRE network managed successfully a cooperative track prosecution task.