Systems of interacting autonomous agents are commonly used for the execution of cooperative tasks. The use of multiple platforms allows for the enhancement of operations performance, robustness and safety. To this end, a key aspect is provided by the inherent redundancy of multi-agent systems.
The project aims at extending the classical control allocation architecture to multi-agent systems, and the main goal is to design a modular optimization policy based on the active exploitation of the different layers of redundancy, i.e. both at input level and at agent level.
The advantages of using a multi-agent framework to perform complex operations has been widely highlighted in the literature, also in connection with the recent advances in AI and learning algorithms. However, a unified setup incorporating the optimized allocation of control, communication, sensing and processing resources together with data-driven mission planning and learning techniques is still missing. This means that, if on the one hand the high potential of multi-agent systems is already well known and recognized, on the other hand such a unified setup may be a significant step towards new technology and innovation frontiers.
The aim of the project is to fill this gap by developing a modular, plug-and-play, cost-efficient operative setup that may be applied to different types of agents, potentially heterogeneous, and for different types of operations and tasks. In addition, data-driven and task-driven design of the system architecture are two key aspects of the innovative vision we endeavor and propose for the analysis of multi-agent operation frameworks.