Modelling adaptive systems using plausible petri nets

One of the main challenges when analyzing and modelling complex systems using Petri nets is to deal with uncertain information, and moreover, to be able to use such uncertainty to dynamically adapt the modelled system to uncertain (changing) contextual conditions. Such self-adaptation relies on some form of learning capability of the Petri net, which can be hardly implemented using the existing Petri net formalisms. This paper shows how uncertainty management and self-adaptation can be achieved naturally using Plausible Petri Nets, a new Petri net paradigm recently developed by the authors. The methodology is exemplified using a case study about railway track asset management, where several track maintenance and inspection activities are modelled jointly with a stochastic track geometry degradation process using a Plausible Petri net. The resulting expert system is shown to be able to autonomously adapt to contextual changes coming from noisy condition monitoring data. This adaptation is carried out taking advantage of a Bayesian updating mechanism which is inherently implemented in the execution semantics of the Plausible Petri net.