Adaptive Fault Tolerance
Adaptive FT in Automotive Applications
Cyber-physical systems (CPS) are engineered systems that are built from and depend upon the synergy of computational and physical components. CPS continues to grow in prevalence and complexity: Life-critical systems such as aircraft, nuclear and chemical reactors, power generation systems, and automobiles are increasingly being controlled by computer. Traditional fault tolerance is expensive in terms of massive redundancy. For automotive embedded systems, the computer systems in power train and chassis domains, as well as certain major tasks for the recent self-driving vehicle (Unmanned Ground Vehicle), such as localization, path planning, etc., require a high level of fault tolerance. Our objective of this paper is to demonstrate while massive hardware redundancy is required, its actual runtime deployment can be done adaptively, based on the current state of the controlled plant. In many, if not most, instances, the plant is in a state which allows for a lowered level of fault-tolerance. Two cases, ABS in a straight line and ESP during a turn, are discussed and the corresponding benefit for overall reliability is also mentioned.