Using Real-World Driving Databases to Generate Driving Cycles With Equivalence Properties

Abstract

Due to the increasing complexity of vehicle design, understanding driver behavior and driving patterns is becoming increasingly more important. Therefore, a large amount of test driving is performed, which together with recordings of normal driving, results in large databases of recorded drives. A fundamental question is how to make best use of these data to devise driving cycles suitable in the development process of vehicles. One way is to generate driving cycles that are representative for the data or for a suitable subset of the data, e.g., regarding geographical location, driving distance, speed range, or many other possible selection variables. Further, to make a fair comparison on two such driving cycles possible, another fundamental requirement is that they should have similar excitation of the vehicle. A key contribution here is an algorithm that combines the two given objectives. A formulation with Markov processes is used to obtain a condensed and effective characterization of the database and to generate candidate driving cycles (CDCs). In addition to that is a method transforming a candidate to an equivalent driving cycle (EqDC) with desired excitation. The method is a general approach but is here based on the components of the mean tractive force (MTF), and this is motivated by a hardware-in-the-loop experiment showing the strong relevance of these MTF components regarding fuel consumption. The result is a new method that combines the generation of driving cycles using real-world driving cycles with the concept of EqDCs.