Edward Lee
Time-sensitive cyber-physical systems integrate an unpredictable physical world, event-triggered and time-triggered actions, communication networks with varying reliability, and computation that takes unpredictable amounts of time. It might seem that we have to abandon any notion of determinism and embrace uncertainty. In this talk, I will argue to the contrary, that deterministic models become even more valuable in such scenarios. Among their many benefits, they enable systematic testing, they shift complexity from application logic to infrastructure, they enable fault detection, they facilitate composability, and more. The key is to understand that determinism is a property of models, not of physical realizations. In engineering, our primary goal is to coerce the physical world to match our models. In contrast, in science, the primary goal is to coerce the models to match the physical world. The engineering approach is much more tractable for complex time-sensitive cyber-physical systems. In this talk, I will examine what we mean by “determinism” in engineering, science, and a bit in philosophy. Whether a model is deterministic or not depends on how one defines the inputs and behavior of the model. I will conclude by outlining a practical deterministic model well suited for designing practical, complex, distributed, time-sensitive cyber-physical systems.
Source
Determinism in Time-Sensitive Cyber-Physical Systems
