Should I wait for supervisory controller synthesis to finish or change the settings?
other
When performing supervisory controller synthesis using the ESCET toolkit, a user has no indication of how long the algorithm will take before it terminates. In fact, a user does not know whether they should wait until it finishes, or if they should try again using different settings. In this master thesis, we explore the symbolic computations used in the synthesis algorithm to determine when a set of settings produces a bad run of the algorithm. In particular, we explore the saturation algorithm used for performing symbolic reachability computations. We consider the algorithm from three perspectives: the input of the algorithm, its inner workings and the implementation of the algorithm. We find that when using bad settings, the algorithm performs a lot of duplicate work, which can be measured during a run. In the end, we come up with a predictor using this data on duplicate work that can detect a bad run when having seen 20% of that run, with 82% accuracy.
TNO Identifier
1030797
Publisher
Radboud University
Collation
63 p.
Place of publication
Nijmegen