The limited possibilities to evaluate the energy eciency of driving algorithms forconnected and autonomous vehicles (CAVs) make it very dicult for policymakers to decideon the potential of autonomous driving. This study is introducing a method to analyze theenergy performance of a driving algorithm under various simulated trac conditions usingthe microscopic trac simulator SUMO. The method can also be used to optimize drivingalgorithm parameters for chosen trac scenarios. Therefore, a tool-chain is developedthat can simulate a CAV under many trac scenarios in SUMO systematically. In thosescenarios, one or more vehicles are controlled by the implemented driving algorithm. Theresulting driving cycles are then analyzed by a forward-facing energy model to calculate theconsumed energy. To validate the model, three measurement cycles under real urban tracconditions were taken and the speed and state of charge (SOC) data of the test vehicle, a2017 Tesla Model S 75D, were collected. The energy model was shown to be highly accurateand the simulated road network and trac, which were chosen to represent the same urbantrac scenario as the measured cycles, were shown to result in similar statistics as themeasurements. A simple driving algorithm that is already implemented in SUMO’s Krauscar-following model was chosen to verify the model’s applicability. For di↵erent values ofthe algorithm parameters acceleration and deceleration, a range of random driving cycleswas simulated. In the simulations and the measurements, the e↵ect of higher and loweruse of auxiliary systems was also analyzed. The results show that the analyzed drivingalgorithm achieves similar results for the energy consumption as the human driver in themeasurements with the best performing parameters. Also, the significance of auxiliarysystem use on the energy consumption and its e↵ect on a driving algorithm’s parameterto remain energy ecient due to the higher impact of the trip duration is pointed out.
Buhk et al. (Thu,) studied this question.