Self-organized Message Scheduling for Asynchronous Distributed Embedded Systems

Abstract

A growing number of control systems are distributed and based on the use of a communication bus. The distributed nodes execute periodic tasks, which access the bus by releasing the messages using a priority-based mechanism with the goal of minimal message response times. Instead of randomly accessing the bus, a dynamic scheduling of messages technique based on adaptation of time offsets between message releases is used. The presented algorithm, called DynOAA, is executing on each node of the distributed system. It takes into account the current traffic on the bus and tries to avoid simultaneous release of messages by different nodes, hence reduces the likelihood of conflicts and need for repeated release. In this paper, we first address single bus (segment) systems and then extend the model and the offset adaptation algorithm to systems that use multiple buses (segments) connected by a communication gateway. A rating function based on the average of maximum response times is used to analyze DynOAA for the case of CAN-bus systems based on bit-accurate simulations. Experiments show the robustness of the algorithm (1) in case of fully asynchronous systems, (2) ability to deal with systems that change their configuration (add or remove message release nodes) dynamically and (3) model systems containing multiple bus segments connected by a gateway. The approach is also applicable to other priority-based bus systems.