ASM, controller synthesis, and complete refinement

While many systems are naturally viewed as the interaction between a controller subsystem and a controlled, or plant subsystem, they are often most easily initially understood and designed monolithically, simply as a collection of variables that represent various aspects of the system, which interact in the most self-evident way. A practical implementation needs to separate controller from plant though. We study the problem of when a monolithic ASM system can be split into controller and plant subsystems along syntactic lines derived from variables' natural affiliations. We give restrictions that enable the split to be carried out cleanly, and we give conditions that ensure that the resulting pair of controller and plant subsystems have the same behaviours as the original design. We relate this phenomenon to the concept of complete refinement in ASM. Making this strategy work effectively, usually requires a nontrivial domain theory, into which a number of properties which are neither the sole possession of the controller subsystem nor of the plant subsystem must be placed. We argue that these properties are latent in the original monolithic model. We illustrate the theory with a case study concerning eating with chopsticks. This leads to an extension of controller synthesis for continuous ASM systems, which are briefly covered. The chopsticks case study is then extended into the continuous sphere.