Feature Interaction Detection in the Automotive Domain

Abstract

The main goal of our research is to develop techniques for the detection of feature interactions for embedded systems in the automotive domain. Automotive systems are cyber-physical systems (CPS), which are composed of a "cyber" part that consists of software components running on digital hardware and a "physical" part that is the mechanical processes monitored by the software. Feature interactions in automotive systems arise from the activation of two or more features sending requests to the mechanical processes that create contradictory physical forces, possibly at distinct times, that cause unsafe behaviours. An example is having simultaneous requests to apply the brakes and the throttle. While both actions may be "correct" according to the intended behaviour of each feature, their interaction is undesired and potentially dangerous. To deal with the feature interaction problem, we propose to perform analysis at design time by using formal methods to detect all possible interactions.