Dynamic Tasking for Satellite Constellation: A Multi-Agent based approach using cooperative auctions

The growing demand for ever-larger satellite constellations poses significant challenges in terms of automation and coordination capabilities. New advances in onboard automation enable satellites to make highly autonomous decisions, providing the basis for complex applications such as dynamic task allocation in a satellite constellation. By coordinating these satellites through a negotiation-based approach, it is possible to improve space mission return and enable a more efficient allocation of complex tasks. This thesis paves the road toward dynamic task allocation using a multi-agent based architecture and cooperative auctions. A constellation composed of autonomous satellites is modeled as a multi-agent system and the problem of allocating new service requests is formulated accordingly. The model comprises satellite's capabilities, on-board resources, and various constraints. Then, an auction algorithm is designed and implemented to enable a cooperation mechanism between satellites and efficiently distribute the incoming requests to speed up on-ground planning. Each agent bids on the announced service requests by combining the contributions of several bidding terms, each one modeling a specific motivator for the satellite (e.g., resource availability or the presence of conflicts). A set of test cases has been designed to ensure the code's integrity and the correctness of the auction execution. A scalability analysis is presented in order to assess the model's performance for an increasing number of satellites interacting dynamically in a constellation. In addition, a complexity analysis is developed to evaluate the time complexity in an environment with many agents, each of which presents its own time-varying characteristics during the auction. Finally, a sensitivity analysis allows fine-tuning the scaling parameters of the functions of each bidding term, as well as the weights that each term should have in the overall bid, with the purpose of evaluating the influence and impact of each component on the overall performance.