Finding representative routes of a vehicle’s operations

Abstract

Vehicles can be tested and assessed in many ways, such as simulation or in real traffic. Regardless of the assessment method, a test route must be defined, and, for the results to be meaningful, it should represent a realistic operating scenario. This thesis presents a method for identifying real-world test routes that are representative of a customer’s road-related operating conditions. Fixed driving cycles are commonly used in vehicle evaluation, but they may fail to capture the variation in road environments caused by customer-specific usage. This can reduce the accuracy of estimates of vehicle performance and energy consumption. To address this, customer log data are matched with road data and used to construct deterministic operating conditions. These are then used to estimate customer-specific stochastic operating condition parameters for road type, speed limits, slopes, curvature, and traffic control devices. Routes are generated from a chosen origin and processed using the same road-data extraction procedure. A representativeness score is then defined as the degree to which the features of each route match the corresponding customer feature distributions. The score enables both generated routes and customer log files to be ranked according to how closely they match the customer’s typical road operating conditions. The results show that the proposed score can identify routes that are more representative of a customer’s road usage, as well as routes corresponding to more unusual missions. Comparisons with estimated specific energy consumption indicate that the representativeness score is related to energy consumption, although it is not designed to optimise for it directly. A regression-based energy prediction model further suggests that the stochastic parameters describing the road contain relevant information for estimating energy consumption. The thesis concludes that customer usage can be described by stochastic operating conditions and that this representation provides information that can support the statistical selection of representative real-world routes.