Consensus-seeking using distributed predictive control
Keywords:Distributed and large-scale systems; Multi-agent systems; Distributed control; Optimal and model predictive control
Description:Engineered systems are becoming increasingly complex and larger in size, which presents a need for the distribution of decision-making processes that interact with or are part of these large-scale technologies and applications (such as autonomous vehicle teams, mobile sensor networks, nodes of a communication network). An important problem that arises among such distributed decision-making systems (often called agents), is related to consensus-seeking and rendezvous, which have been studied in various fields such as computer science and communication networks. The problem becomes even more interesting when agents are modeled as constrained dynamical systems and the interconnection topologies are time-varying. In this setting, the consensus-seeking and rendezvous problem consists of designing distributed control strategies such that the state or output of a group of agents asymptotically converge to a common value, a consensus point, which is agreed upon either a priori or on-the-fly using some negotiation scheme. We may assume that a consensus point is not fixed in advance, but is rather determined by an optimal control problem, which is solved in a distributed way. Thus, a combination of distributed controller design and consensus seeking algorithms is required.
This MSc project focuses on studying the interplay between distributed model predictive controllers and various negotiation schemes of optimal consensus (e.g. incremental subgradient-based), and investigate conditions for asymptotic convergence of such distributed control schemes. The proposed approach can be tested in optimal synchronization problems involving constrained subsystem dynamics, such as multi-vehicle coordination or oscillator networks.