On-road Brain–Body–Vehicle multimodal dataset for four-level road-rage.

We design and conduct a near-naturalistic on-road Brain–Body–Vehicle (BBV) protocol to synchronously collect EEG, peripheral physiological signals, and CAN-bus vehicle dynamics, and establish a four-level road-rage multimodal dataset under real-vehicle driving conditions. For convenience, two representative elicitation paradigms are denoted as Experiment A and Experiment B, respectively. The primary trigger of Experiment A is longitudinal car-following distance threat, induced by repeated frustration-driven acceleration and deceleration pulses; the primary trigger of Experiment B is lateral spatial intrusion, elicited by discretionary lane-change overtaking together with high-acceleration right- and left-turn maneuvers. During each run, participants continuously reported their road-rage intensity via a button-based input; the online labels were then further reviewed and calibrated through a video-assisted verification procedure to ensure temporal precision and cross-subject consistency. This dataset links cortical activity, autonomic arousal, and vehicle-control dynamics in authentic traffic conflicts, providing a practical foundation for BBV modeling and for studying the transfer from high-cost multimodal learning to CAN-only deployable inference.(Due to ethical and privacy requirements, the full dataset is not publicly released at this stage; we only make a representative sample from one participant available to support methodological reproducibility and illustrate the data structure, while access to the remaining data may be granted under controlled conditions upon reasonable request and compliant approval.)