Knowledge and communication in decentralized discrete-event control
Karen Rudie, S.L. Ricker · 2000
A formal method for reasoning about knowledge in distributed systems is applied to the analysis of decentralized discrete-event control problems. Solutions to this class of control problems require that controllers achieve their control objectives without communication. A necessary and sufficient condition is given (equivalent to one from existing discrete-event control theory) to describe when decentralized controllers have enough knowledge to find a control solution. When controllers do not have sufficient knowledge, a solution where controllers may communicate is presented. The relationship between communication and control is difficult. Control decisions may be affected by information a supervisor received from another supervisor. The content of the information that is communicated could be affected by information the communicating supervisor previously received. Procedures are derived for incorporating communication into decentralized discrete-event control. These procedures yield a control solution while ensuring that supervisors communicate in a consistent manner.