On Modularity in Reactive Control Architectures, with an Application to Formal Verification

Modularity is a central principle throughout the design process for cyber-physical systems. reduces complexity and increases reuse of behavior. In this article we pose answer following question: how can identify independent “modules” within structure reactive control architectures? To end, propose graph-structured architecture call decision show it generalizes some architectures that are popular in Artificial Intelligence (AI) robotics, specifically Teleo-Reactive programs (TRs), Decision Trees (DTs), Behavior (BTs), Generalised ( k -BTs). Inspired by definition module graph theory [ 16 ] define modules structures each possesses canonical decomposition into its modules, which be found polynomial time. We establish intuitive connections between our proposed modularity structured programming. BTs, -BTs, DTs one-to-one correspondence with their subtrees. naturally characterize DTs, TRs properties decomposition. This allows us to recognize equivalent these quadratic Following McCabe 26 ], measure called essential on structures, measures degree they decomposed simpler modules. -BTs as unit-essential complexity. Our concept extends formal verification, under any verification scheme capable verifying structure. Namely, prove modification has no greater flow-on effects than an individual action enables done locally hierarchically, where verified then repeatedly modified, replaced while preserving correctness. illustrate findings, present example solar-powered drone completing reconnaissance-based mission using use Linear Temporal Logic-based verify correctness one modify correctness, considering only those have been modified.