Towards Response-Time Analysis of Complex Real-Time Systems by using ParametricWorst-Case Execution-Time Estimate on Tasks – A Case Study for Robotic Control System

Avoiding timing-related errors in complex industrial real-time software systems becomes more and more important. In our target domain, such complex software systems are developed in C and realized with periodic tasks, executed on a single processor under fixed-priority preemptive scheduling. The tasks exhibit intrinsic dependencies with respect to execution times; often the execution time of a task is dependent from inter-process communication or globally shared variables. Thus, the execution time is highly variable and using a pessimistic, safe, upper bound on the execution time (worst-case execution time, WCET) in analysis of the system would give unacceptable pessimistic results. However, existing techniques for schedulability analysis, such as the well known Response-Time Analysis (RTA), typically use WCET as their input. In this paper, we illustrate the problem by using a timing model inspired by a robotic control system from ABB. We show how models of tasks whose execution time is dependent on asynchronous message-passing and globally shared state variables, and how we can model this execution time as a parametric WCET (PWCET). Further, we show how we, in this example can use TIMES to formally derive the parameters for the PWCET in order to obtain a concrete WCET to be used in the RTA.