AUTOSAR Runnable Scheduling for Optimal Tradeoff between Control Performance and CPU Utilization
Tae-Wook Kim, Gun-Min Lee, Jong-Chan Kim · International Conference on Control, Automation and Systems · 2018
Automobile control systems should be designed for optimal control performance through efficiently utilizing its underlying hardware resources like CPU. For this purpose, lots of real-time scheduling and system optimization methods have been developed. However, most of these researches assume that the unit of software is a real-time task and its period and deadline is freely adjustable. However, in practice, AUTOSAR, the de-facto standard automobile software architecture, organizes a control system as a collection of runnables, which is a much smaller unit of functions compared to task. Moreover, a sensing-to-actuation control loop is composed of a chain of runnables, which makes end-to-end delay analysis more complex. With this motivation, this paper formulates a runnable scheduling problem as an integer programming finding optimal runnable periods that maximizes the control performance with a given linear-approximate control cost function.