Design and performance analysis of fault tolerant TTCAN systems.

摘要

Continuous demand for fuel efficiency mandate "Drive-by-Wire" systems. The goal of Drive-by-Wire is to replace nearly every automotive hydraulic/mechanical system with electronics. Drive-by-Wire and active collision avoidance systems need fault tolerant networks with time triggered protocols, to guarantee deterministic latencies. CAN is an event triggered protocol which has features like high bandwidth, error detection, fault confinement and collision avoidance based on message priority. However, CAN do not ensure message latency, which is critical for real time application. TTCAN (Time Triggered CAN) removes this fallacy of CAN by providing exclusive time windows for those messages that need deterministic latencies. In addition to the exclusive windows, there are arbitration windows too, which make way for event triggered communications. In TTCAN, if an error occurs within an exclusive or arbitration window, retransmission of the message is not allowed. If the message that encountered the error is a safety critical message, then the transmission error can compromise the safety of the vehicles. The thesis work presented here proposes three techniques to increase fault tolerance of TTCAN systems. The proposed techniques increase fault tolerance of TTCAN systems by improving system matrix design and incorporating redundant bus. A detailed description of architectures and algorithms required to implement these techniques has been presented. These techniques have been studied analytically and by using simulation. The results show significant improvement in the fault tolerance of TTCAN.