Error recovery in time-triggered communication systems using servers

Abstract

In communication systems, transient faults will eventually occur. Thus, some mechanism is necessary to handle them and achieve appropriate levels of reliability, particularly in safety-critical systems. One possibility is to rely on temporal redundancy, i.e., using message retransmissions. General requirements for such a mechanism would include a parsimonious use of extra bandwidth while guaranteeing the schedulability of the message set. In this paper we propose using on-line traffic scheduling together with scheduling servers to recover message errors in time-triggered systems on Controller Area Network (CAN), taking advantage of the Flexible Time-Triggered CAN protocol. This novel mechanism is shown to offer a desired error recovery latency using much less extra bandwidth than typical approaches used in time-triggered systems. In this paper we present this novel error recovery mechanism, including a thorough characterization as well as configuration guidelines, namely concerning how to choose the server parameters (type, period and capacity). The correctness of the proposed approach and its superior performance are validated with simulation using several communication benchmarks available in the literature.