Embedded Files
Monograph
"Supervisor Localization: A Top-Down Approach to Distributed Control of Discrete-Event Systems", Lecture Notes in Control and Information Sciences, vol. 459, Springer, 2016.
Selected Recent Papers
Supervisor localization of discrete- A distributed algorithm for resource Relative observability and coobservability
event systems under partial allocation over dynamic digraphs, of timed discrete-event systems,
observation, Automatica, 2017 IEEE Trans. Signal Processing, 2017 IEEE Trans. Automatic Control, 2016
Projects
1. Supervisor localization: a top-down approach to distributed control of discrete-event systems
We address this question: Given a collection of independent agents as the plant and some desired collective behavior as the specification, what should individual agents do (sensing and decision making) so as to enforce the specification, and realize control performance identical to that achieved by the global monolithic supervisor?
We propose a top-down approach that systematically decomposes a synthesized supervisor into local controllers, one for each individual agent; the collective behavior of the local controllers are proved to be identical to the global optimal and nonblocking controlled behavior. The result is a purely distributed control architecture in which each agent is controlled by its own local controller, while communicating (via event synchronization) with its nearest peers. Read more.
2. Surplus-based average consensus in networked multi-agent systems
We address this question: Given a network of agents (single integrators) with a directed communication topology, can we design a distributed algorithm for the agents to achieve average consensus over (arbitrary) strongly connected digraphs?
We propose a distributed algorithm, augmenting for each agent a surplus variable to keep track of local state change. Each agent updates its state according to neighbors' state and surplus information. This strategy is proved to achieve average consensus over any strongly connected digraph, in particular without the balanced condition (equivalently column-stochastic adjacent matrix). Read more.
3. Relative observability of discrete-event systems under partial observation
We address this question: In supervisory control of discrete-event systems under partial (event) observation, is there a language property that is stronger than observability, weaker than normality, and closed under arbitrary set unions, so that the corresponding supremal sublanguage exists for supervisor synthesis?
We propose relative observability, with a fixed ambient language relative to which the standard observability is tested. Relative observability is proved to be stronger than observability, weaker than normality, and closed under arbitrary set unions. We design an algorithm that computes the supremal relatively observable (and controllable) sublanguage of a given language. This synthesized supremal sublanguage is the new supervisor under partial observation, with generally more permissive controlled behavior than the normality counterpart. Read more.
Funding:
- MEXT Program to Disseminate Tenure Tracking System, 2014-2019.
- "Robust and adaptive distributed control of discrete-event systems", JSPS Grant-in-Aid for Young Scientists (B), 2016-2019.
- "Internal-model distributed control design for complex networks of dynamic agents", JSPS Grant-in-Aid for Young Scientists (B), no. 26870169, 2014-2015.
Page updated
Google Sites
Report abuse