Optimization of feature interactions for automotive combustion engines

With increasing numbers of features in automotive systems, feature interaction (FI) becomes more and more relevant regarding safety and emissions. Hence, undesired and potentially even unknown FIs pose a major challenge now to the automotive industry, but optimizing desired FI may even help to further reduce CO2 emissions. We propose a new approach for optimization of feature interactions at run-time for automotive combustion engines. It integrates an optimization objective (minimize CO2 emission) with both soft and hard constraints (e.g., related to certain temperatures). In the course of iterations of hill-climbing optimization at run-time, the integrating objective function is dynamically adapted for heuristic coordination of FIs. While our approach with its linear time and space complexity does not opt for finding globally optimal solutions, it coordinates feature interactions, both desired and undesired ones. Safety-critical problems are avoided using hard constraints and via release/inhibit conditions for features. Our prototypical implementation of this approach showed promising and realistic results in a simulation environment.