Distributed Sense and Control Systems
Substantial research challenges exist in the design and verification of large-scale, complex, distributed sensing, actuation and control systems. Three topics of particular interest are the design in information flows, cooperative behavior between distributed agents, and verification of distributed, asynchronous control systems. Each of these topics relates to a difficult aspect in the proper operation of large-scale distributed system in which the temporal scales of the underlying dynamics of the systems, the rate of communications between agents, and the latency in computation and multi-threaded execution cannot be separated. The additional need to be able to rapidly design, implement and commission such systems requires new techniques in modeling, analysis, design and verification.
To approach the interlinked challenges of information flow, cooperative behavior and verification, we plan to combine tools and recent advances from control theory, networked systems and computer science. The primary tools that we expect to build on are graph theory, partial order theory, temporal logic, graph grammars, formal methods and optimization-based control. This combination of tools allows us to model and analyze complex, protocol-based control systems by using temporal logic to specify desired behavior, graph theory (in particular the graph Laplacian and other associated matrices) to model and design the information flow, graph grammars to design cooperative behavior, lattice theory and Lyapunov theory to understand convergence properties via invariant sets, and model-checking and receding horizon control to design systems whose asynchronous execution sequences satisfy a given specification. Previous results in each of these areas has demonstrated the efficacy of modeling and analysis of distributed, asynchronous sensing, actuation and control systems; future work will focus on advances required to support large-scale systems and modularity.
- Develop algorithms, protocols and verification tools for distributed control that are aimed at optimized resource allocation between multiple agents in a decentralized and scaleable manner.
- Extend existing work in gossip algorithms, load balancing and consensus protocols to allow more sophisticated scheduling of shared resources in a dynamic environment with model-based predictions of future demand/availability.
- Application areas that will be used as drivers for the theoretical advancements include (1) power management networks in handheld devices, vehicles, buildings, manufacturing plants and geographic regions; (2) aircraft avionics and (3) cooperative control of multi-vehicle systems.