Taming confusion for modeling and implementing probabilistic concurrent systems

In concurrent systems, the choice of executing the next transition depends both on the timing between the agents that make independent or collaborative interactions available, and on the conflicts (nondeterministic choices) with other transitions. This creates a challenging modeling and implementation problem. When the system needs to make also probabilistic choices, the situation becomes even more complicated. We use the model of Petri nets to demonstrate the modeling and implementation problem. The proposed solution involves adding sequential observers called agents to the Petri net structure. Distributed probabilistic choices are facilitated in the presence of concurrency and nondeterminism, by selecting agents that make the choices, while guaranteeing that their view is temporarily stable. We provide a distributed scheduling algorithm for implementing a system that allows distributed probabilistic choice.