Hierarchical Agent-based Command and Control (C2) Systems

[bibshow file=MyPublications.bib]

c2    Cooperative missions using a team of autonomous vehicles poses a challenge to the designers of the underlying C2 system. Not only does the C2 system have to handle its own mission tasks and onboard sensor data, but has to deal with potentially complex interactions among the peer vehicles in the team. One common approach to ease the problem is to adopt the divide-and-conquer model where high level mission tasks are decoupled from the low level vehicle navigational tasks. The division provides a clear view on the overall control architecture and makes the C2 system development and maintenance more manageable. Typically the vehicle navigational control remains unchanged regardless of the complexity of the overall mission objectives. However, the requirements of the high level mission tasks evolve with the nature of the cooperative missions specified for a team of vehicles. The underlying C2 system must be extensible to handle the new requirements as they are introduced.

    In [bibcite key=Tan:2012southampton,5547041,Tan2009a], we adopted a multi-agent approach in the C2 system where mission, navigation and vehicle control tasks are allocated to individual software agents that are arranged in a hierarchical order according to their corresponding control responsibilities. To allow for different mission behaviors and cater to various payload modules with potentially different mission requirements, we adopted a Backseat Driver (BD) paradigm where mission decisions are made based on the input provided by a pool of BD agents. This pool is termed as Agent Services (AS). Each BD agent in the AS implements different algorithms and monitors various sensor data to generate inputs in the form of mission points, which when accepted by the Captain agent for execution, achieve a specific mission task. The interaction between the Captain agent and the AS ensures mission objectives are achieved. The resultant C2 system has been successfully used in numerous single and multiple vehicle missions: cooperative localization [bibcite key=starfishJournal,tanoceans2011,Tan:2012Baltimore,BathyAuRo_2014,Tan:JOE2012], adaptive sampling, source localization with robotic swam [bibcite key=BIOCAST] and adapted to allow interoperability with MOOS for a field experiment involving STARFISH AUV and MIT USV [bibcite key=STARFISH].