Simulation-Driven Optimization of Real-Time Control Tasks

Abstract

In this paper we define a simulation-driven process to improve the design of real-time control systems. The process aims at exploring the interplay betwen control performance and real-time behavior of control tasks. The traditional design flows based on the definition of implicit tasks deadlines on control functions are extended to include the exploration of relaxed deadlines and order of execution constraints. Relaxed deadlines, coupled with an optimization approach to find feasible task sets, allow the exploration and evaluation of different task implementations. The definition of relaxed deadlines and the evaluation of task implementations is performed using the T-Res (Time and Resource) scheduling simulation framework [21] under Simulink. The problem is defined as a quadratic optimization problem using a tight upper bound formulation of the task response times. The application of the method to a quadcopter case study shows how the consideration of the control performance in the definition of the timing parameters of interest can lead to an improved design.