I present a computational architecture designed to capture certain properties essential to actions, including compositionality, concurrency, quick reactions, and resilience in the face of unexpected events. It uses a structured internal state model and complex inference about the environment to inform decision-making. The properties above are achieved by combining interacting procedural and probabilistic representations, so that the structure of actions is captured by Petri Nets, which are informed by, and affect, a model of the world represented as a Probabilistic Relational Model. I give both a theoretical analysis of the architecture and a demonstration of its use in a simulated robotic environment.