Cyber-physical systems involve hardware and software that are highly interconnected
and complex. Unexpected failures in these systems can cause material damages, cost
a lot of money and reputation, and can risk human lives too. The correct way of
avoiding unexpected failures is to make a mathematical model of the system and
to perform exhaustive analysis on it. Also, mathematical models are inevitable for
performance analysis of the systems (e.g., finding throughput and bottlenecks) and
to plan for extensions (e.g., measuring cost over performance factor).
Petri net, since its inception in 1962, has been used for modeling and performance
analysis of discrete event systems. In the 1970s and 80s, Petri net was tried out
on modeling smaller discrete manufacturing systems like flexible manufacturing
systems with a handful of CNC machines and robots. Very soon, it became apparent
that even for smaller systems, the Petri net models become huge. Also, the state
spaces generated from the models become too big to handle (the problem of state
explosion). Of course, for a large Petri net model of a real-life discrete system, the
state space is enormous, and analysis of it is not a possibility.
Also, simulations take too much time, as the simulation of a Petri net means
tracking the movement of tokens as they pass through places and transitions. For a
large Petri net with hundreds of places and transitions, monitoring the movements
of all those tokens makes the simulations slower. The slow simulations also make
Petri net not suitable for real-time control. If faster execution of Petri net is possible,
then Petri net could become an ideal candidate for supervisory control of real-time
systems