Computationally efficient supervisor design for discrete-event systems

  • Authors:
  • Lei Feng

  • Affiliations:
  • University of Toronto (Canada)

  • Venue:
  • Computationally efficient supervisor design for discrete-event systems
  • Year:
  • 2007

Quantified Score

Hi-index 0.00

Visualization

Abstract

In response to the exponential computational complexity of supervisor design for discrete- event systems and the trend towards distributed control in applications, this dissertation extends current results towards a unified comprehensive theory of control architecture, enabling the design of decentralized controls with feasible computational effort. Major contributions include: (1) identification of universal primitive components, such as 'servers' and 'buffers' in a manufacturing (or similar) system, whose control actions are determined locally and which make up systems more amenable to nonblocking verification than general structures; (2) creation of a new modeling tool, control-flow net, that can exhibit the interconnection among plant components and specification via a graph, so that a system can be simplified and decomposed according to the net structure; (3) specification of the causal relation between system blocking and special net structure in the control-flow net, which enables the qualitative derivation of control logic independent of system numerical parameters; (4) a flexible decentralized and hierarchical supervisory control architecture based on control-flow net to reduce computational effort; and (5) application and extension of the 'observer' concept to constructing abstractions used in hierarchical control, so that the nonblocking property is strictly preserved. In the context of supervisory control theory, this dissertation has mathematically validated these 'common-sense' principles and confirmed their effectiveness through several examples.