Closing the gap in control system implementations

摘要: A cyber-physical system tightly coordinates discrete computation and continuous control of physical resources. Most safety-critical systems run sophisticated algorithms at their core. While these have been studied thoroughly from a theoretical standpoint, implementation on real platforms raises many issues that need to be addressed ensure the reliability systems. theory studies properties based analysis mostly ignores effect implementing laws in software, effects can significant impact performance or stability law. Consider following mismatches between assumptions made by guarantees provided platforms. First, theoretic methods are analysis, which assume numbers infinite-precision. However, software implementation, mathematical approximated using floating-point fixed-point arithmetic. Second, time is often assumed negligible, may not case implementations. Third, when multiple applications shared platform, schedulability requirements corresponding tasks impose additional constraints behavior systems, traditional design do take into account. In this thesis, we address three problems related controller for show how property verified under action synthesize minimize quantization error quality. naive some infeasible due required provide memoization scheme feasibility along with maintaining expected performance. problem scheduling single processor static dynamic scheduler synthesis strategies maintain achieve optimal Solving takes an important step towards closing gap