Architecture for safety-critical transportation systems

In many industrial systems, including transportation, fault tolerance is a key requirement. Usually, fault tolerance is achieved by redundancy, where replication of critical components is used. In the case of transportation computing systems, this redundancy starts with the processing element. In this paper, we use Markov models to assess the level of safety with different redundancy techniques used in the literature. More specifically, we give implementation details for various architecture options and evaluate one out of two (1oo2) and two out of three (2oo3) implementations. We observe that both 1oo2 and 2oo3 can reduce the average probability of failure per hour (PFH) down to 10-7 which provides Level-3 (SIL3) safety according to the standards.