Technical Description of the Project
We focused our research on the realm of cooperative multi-agent systems, by which we mean a number of homogeneous and/or heterogeneous agents collaborating autonomously through spatially-distributed physical interactions and wireless communication channels to jointly accomplish certain desirable objectives. More specifically, we develop a formal design framework for cooperative multi-agent systems, which synthesizes individual control policies for each agent as well as coordination strategies among them such that the joint efforts of the agent result in satisfaction of a formal global performance specification. More specifically, by characterizing the collective behavior emerged through the cooperation of agents as languages generated by a discrete-event system (DES), the proposed automatic synthesis framework solves the coordination and control problems of cooperative multi-agent systems via \divide-and-conquer": (I) first the global specification is decomposed into local subtasks according to each agent's sensing and actuating capabilities; (II) a learning-based synthesis algorithm is developed to synthesize a local supervisor for each agent to fulfill the local task; (III) a compositional verification procedure is deployed to justify whether or not the joint behavior of all the controlled agents shall satisfy the global specification. Our approach provides a learning-based method to automatically synthesize coordination and control strategies for cooperative multi-agent systems such that a global specification can be met, even under the circumstance that the DES model of each agent is not given a priori. In addition to the theoretical contributions, computational and software tools are also developed to incorporate automatic supervisor synthesis with inter-robot communication in an effort to implement the framework. A demonstration example of multi- robot coordination to achieve a request-response team specification is presented to justify the effectiveness of the proposed framework.
This work has been partially funded by the following National Science Foundation (NSF) grant CPS: TTP Option: Synergy: Collaborative Research: Dependable Multi-Robot Cooperative Tasking in Uncertain and Dynamic Environments
Significance of the Work
Our research effort centers on the fundamental question essential for building distributed and cooperative systems that can function robustly and reliably in unknown and dynamic environments. Focusing on cooperative multi-agent teams, the central novelty of our research is in developing the foundations of a scalable, reliable and provably correct formal design theory. The proposed design theory will guarantee a given global performance of multi-robot teams through designing local coordination rules and control laws.
Papers related to this project
- Dai, A. Benini, H. Lin, P. J. Antsaklis, M. J. Rutherford and K. P. Valavanis, “Learning-based formal synthesis of cooperative multi-agent systems with an application to robotic coordination.”, 2016 24th Mediterranean Conference on Control and Automation (MED), Athens, 2016, pp. 1008-1013. doi: 10.1109/MED.2016.7536071.
- J Dai, A Benini, H Lin, PJ Antsaklis, MJ Rutherford, KP Valavanis, "Learning-based Formal Synthesis of Cooperative Multi-agent Systems", arXiv preprint arXiv:1705.10427