Characteristic map based idle speed control of a spark ignition engine

In this contribution a novel off-line parametrization method of a characteristic map based idle speed control strategy for a spark ignition engine is proposed. The new startegy uses a low order system air-path model of a spark ignition engine (see[1]), which represents the dominating dynamics. The characteristic map is described in an adequate way for the use of an optimization routine to find the locally optimal parameters corresponding to the grid points of the characteristic map. One of the major challenges is the reduction of complexity by still achieving performance benefits and maintaining stability. This will be done by using linear PI controllers spanning the search area. The optimization is carried out with continuous action reinforcement learning automata (CARLA, [2]), a machine learning algorithm, modified for selecting actions in a continuous space and minimizing the well-know IAE cost functional. The resulting off-line design will be tested against existing idle speed controller realizations and the usefulness of this new strategy will be demonstrated.