A Large-Scale Electric Bus Dataset for Understanding Electric Vehicle Behaviors in Real-world

Abstract: Unlike internal combustion engine (ICE) vehicles, electric vehicles (EVs) have distinct driving dynamics. EVs exhibit faster acceleration due to their instant torque and more efficient drivetrains, while their slower deceleration and longer time gaps are influenced by the use of regenerative braking. Current EV modeling studies often rely on traditional traffic flow models originally designed for ICE vehicles, which fail to accurately capture the unique behaviors of EVs. One of the key challenges in EV modeling is the use of isolated datasets, which limits scalability and real-world applicability. Many EV datasets are typically collected from individual buses or vehicles using CANBus or GPS, but these datasets often lack information about surrounding traffic. Other datasets are gathered in controlled environments, where a few EVs drive in platoons, failing to reflect the complexity of real-world traffic conditions. This study addresses these limitations by integrating multi-source data from the university’s bus fleet, which consists of approximately 30 electric buses and 40 ICE buses. Each bus is equipped with GPS for tracking position and speed, and five cameras (one front, one rear, one left, and two right) to capture the reactions of surrounding traffic. These buses operate across diverse real-world scenarios, including interactions with other vehicles, vulnerable road users, traffic signals, bus stops, and varying weather conditions. To analyze the data, advanced AI techniques are employed, such as MMDetection for object detection and Vision Transformer (ViT)-based single-camera distance estimation to extract trajectories of surrounding objects. After trajectory data cleaning, denoising, and empirical analysis, a dataset in a format similar to NGSIM is generated. This dataset offers a richer understanding of EV behavior in real-world settings, improving EV simulations and contributing to the development of safer and more efficient E-mobility solutions.