Observe locally, control globally

Contact: Stefan Schwoon, Stefan Haar

Supervisory control of partially observable Petri nets has been studied int he literature mostly under a sequential and centralized perspective. With new application challenges from systems biology, it is time to to create a frameworkfor supervisory control of partially observable Petri nets, under the constraints of concurrency of the supervised processes (which are intrinsically non-sequential)and decentralization of observation and intervention.

The objectives targeted by the control must include reachability objectives in safe Petri nets; ideally, the results exceed this domain, to include reprogrammingof long-term behaviors of models arising in cell regulation. In this setting, thestakes are to steer the cell fate into a desired attractor or phenotype, whilestaying clear of fatal pathways (cancerous mutation, etc.)

The strengths of the well-known unfolding technique for safe Petri nets are expected to be instrumental in this work. The controller’s knowledge is a decentralized and asynchronous observation coming from a distributed set of observers. The tasks include:

  • establish the semantics of supervisory control under these circumstances;
  • determine an adequate game theoretic setting, and
  • develop efficient techniques for synthezing controllers.