Autonomous Planning for Surface Engagement using a Multi-Agent System

Less personnel and the introduction of more complex systems on board of future navy vessels requires innovative systems to support the crew in making optimal command decisions. These decisions should result in an allocation of available resources that maximizes the probability to achieve the set of mission goals, provided the external and internal picture. The process of planning tasks and resources across subsystems is challenging, partly due to the heterogeneous nature of these, often proprietary, subsystems and the inter-dependencies between tasks and resource requirements. Currently, the coordination of systems-of-systems on board of navy vessels is predominantly performed by the human crew. This article presents the use of the multi-agent paradigm for supporting the crew by automatically allocating the set of resources that are available on board a vessel to the vessel’s systems. The presented solution leaves the crew in the loop as the ultimate decision maker but facilitates their decisions by providing pre-processed options void of unnecessary details. The solution works as follows: multiple goal-driven software agents cooperate to plan tasks and allocate resources. A single agent can be responsible for vessel functionalities such as sensor–, engagement–, and mobility–management. These agents collaborate autonomously with each other and under dynamically changing external and internal conditions, changing tactics, mission goals and resource availability. Agent decisions are driven by the set of mission goals expressed in a mathematical formulation. This ensures that system resources and assets are allocated in accordance with the mission without increasing the workload of the crew. Conflict resolution resulting from resource usage by agents is handled using a negotiation protocol. The feasibility of the taken approach is demonstrated within an example scenario, where a vessel escorts a high-value asset in enemy territory. This article shows that with the multi-agent paradigm, the described conditions, and in the face of conflicting preferences, the most effective combination of actions is selected and planned.