PDB: Not All Drivers Are the Same -- A Personalized Dataset for Understanding Driving Behavior

Chuheng Wei; Ziye Qin; Siyan Li; Ziyan Zhang; Xuanpeng Zhao; Amr Abdelraouf; Rohit Gupta; Kyungtae Han

cs.CVMar 9, 2025

PDB: Not All Drivers Are the Same -- A Personalized Dataset for Understanding Driving Behavior

Authors: Chuheng Wei, Ziye Qin, Siyan Li, Ziyan Zhang, Xuanpeng Zhao, Amr Abdelraouf, Rohit Gupta, Kyungtae Han,

Organizations: University of California, Riverside · Southwest Jiaotong University · InfoTech Labs, Toyota Motor North America

Abstract

Driving behavior varies across individuals, yet most existing driving datasets do not preserve driver identity or control for factors that may affect driving style. We introduce the Personalized Driving Behavior (PDB) dataset, a multimodal dataset collected from 9 drivers using the same vehicle and route under comparable conditions. PDB includes data from a 128-line LiDAR, external and driver-facing cameras, GNSS, IMU, CAN-bus signals, and heart-rate measurements, with 111,060 frames of LiDAR point cloud data and more than 1.33 million images. Based on these recordings, we construct PDB-Traj, a personalized trajectory prediction benchmark with 1,309 10-second trajectory segments sampled at 10 Hz. Each segment includes the ego trajectory, surrounding-agent trajectories, and driver identity. We further introduce Gated Selective Personalization (Gate-SP), a driver-aware baseline that selectively incorporates driver identity into trajectory prediction. Experiments on PDB-Traj provide reference results for personalized motion forecasting and show that driver information can improve prediction when applied selectively. PDB and PDB-Traj provide a benchmark for studying driver-specific behavior and personalized trajectory prediction.

Figures & tables

Figure 1: The overview of the PDB dataset. ∗ To ensure protect privacy, details regarding driver identities have been withheld.

Dataset	Camera	LiDAR	GPS	CAN-bus	Driver’s Behavior	Physiological Signal	Region/ Platform
KITTI [ 10 ]	$\checkmark$	$\checkmark$	$\checkmark$	$\times$	$\times$	$\times$	Germany
Brain4Cars [ 16 ]	$\checkmark$	$\times$	$\checkmark$	$\times$	$\checkmark$	$\times$	USA
ApolloScape [ 15 ]	$\checkmark$	$\checkmark$	$\checkmark$	$\times$	$\times$	$\times$	China
HDD [ 33 ]	$\checkmark$	$\checkmark$	$\checkmark$	$\checkmark$	$\times$	$\times$	USA
Argoverse [ 7 ]	$\checkmark$	$\checkmark$	$\checkmark$	$\times$	$\times$	$\times$	USA
Lyft Level 5 [ 18 ]	$\checkmark$	$\checkmark$	$\checkmark$	$\times$	$\times$	$\times$	USA

Table 1: A comparison of autonomous driving datasets for multi-sensor fusion.

Figure 2: The PDB data collection sensor setup.

Figure 3: Time synchronization of the multimodal PDB data streams. GNSS provides the GPS time reference for the auxiliary logger, whose clock is used to synchronize the ROS 2 acquisition computer. Driver heart-rate measurements are independently GPS-timestamped and aligned to the common timeline during post-processing.

Data Type	Sensor Model	Frequency (Hz)
Point Cloud	OUSTER OS1-128 LiDAR	10
Raw Images (External and Internal Cameras)	ZED 2i Stereo Cameras	30
GNSS Position	OXTS AV200 GNSS Unit	100
CAN-Bus Signals	OBD-II Interface	SPEED: 83.33 † STEER_ANGLE: 83.33 † GAS_PEDAL: 31.25 † BRAKE_PRESSURE: 41.67 †
IMU	ZED 2i Stereo Camera	400
Driver Heart Rate	Apple Watch Ultra 2	0.2 on average

Table 2: Sensor specifications and sampling frequencies.

Figure 4: Overview of Gate-SP. Ego history and vector-map context are first encoded to produce base and map-aware rollouts, which are fused through a horizon blend for base prediction. Driver-specific information is then selectively incorporated through a turn gate and a trust scalar to form a personalized rollout, which is blended with the base rollout to produce the final Gate-SP prediction.

Figure 5: Qualitative trajectory prediction results on PDB-Traj. Three representative cases from different drivers are shown. Panels (a-1), (b-1), and (c-1) use 3 s observation and 3 s prediction, while (a-2), (b-2), and (c-2) use 3 s observation and 5 s prediction. Gray points denote the observed trajectory, and the green curve denotes the ground truth. CV, LSTM, Base, and Gate-SP predictions are shown for comparison.

	3s $\rightarrow$ 3s			3s $\rightarrow$ 5s
Method	ADE	FDE	balFDE	ADE	FDE	balFDE
CV [ 36 ]	1.59	4.05	5.17	3.91	10.23	11.62
CTRV [ 36 ]	1.36	3.53	3.55	3.46	9.40	9.48
LSTM [ 12 , 1 ]	1.00	2.55	2.79	2.53	6.85	7.29
Transformer (TF) [ 40 ]	2.43	5.08	5.69	4.73	11.13	11.99
TF-DE	1.81	4.28	4.58	4.01	9.82	10.24

Table 3: Benchmark results on the PDB-Traj test split (meters, lower is better). Multimodal methods are scored by their most-likely mode. Base is the map-conditioned predictor without personalization; Gate-SP adds the selective identity conditioning.

Autonomous Driving Datasets

Autonomous driving datasets collect and organize driving-scene data to train and evaluate perception, prediction, planning, and safety systems.

See all 141 papers

PersonaDrive: Controllable Trajectory Prediction with Multi-Dimensional Driving Personas
Aug 15, 2026Chan Lee, Kimin Yun, Yuseok Bae +2Multimodal Trajectory Prediction Autonomous Driving Datasets
Kyung Hee University, Yong-in, South Korea · ETRI, Daejeon, South Korea
DriveDNA: A Large-Scale Multimodal Naturalistic Driving Dataset and Benchmark for Driving Style Identification
Jul 26, 2026Yuhang Wang, Lingyao Li, Hao ZhouAutonomous Driving Datasets Autonomous Driving Benchmarks
University of South Florida · Tampa, Florida, USA · University of Arizona +1
DriveMotion: A Large-Scale Multi-Source Benchmark for Driver Motion Sequence Modeling and Forecasting
Sep 8, 2026Yuhang Wang, Chuheng Wei, Jingxin Yang +2Autonomous Driving Datasets Autonomous Driving Benchmarks
University of South Florida · Purdue University · Stanford University +1

Aug 15, 2026J/K move · Enter open · S save

PersonaDrive: Controllable Trajectory Prediction with Multi-Dimensional Driving Personas

Chan Lee, Kimin Yun, Yuseok Bae, Seong Tae Kim+1

Kyung Hee University, Yong-in, South Korea · ETRI, Daejeon, South Korea

Multimodal Trajectory Prediction Autonomous Driving Datasets

Open paper AskPDF arXiv

Although recent trajectory prediction and end-to-end autonomous driving methods improve robustness in urban environments, they still lack meaningful controllability. Existing benchmarks either provide no persona-conditioned annotations or support only a single urgency spectrum (i.e., emergency, normal, relaxed), which cannot distinguish personas that share the same urgency level but require different driving dynamics. To address this, we propose (i) the Persona-Conditioned Trajectory (PCT) dataset, which decomposes driving personas along two axes, Temporal Urgency and Ride Comfort, and combines three levels of each to form a grid of nine personas, each paired with natural-language descriptions and trajectories, and (ii) PersonaDrive, a framework that can learn driving personas from language and can generate persona-specific trajectories. PersonaDrive incorporates Persona-Conditioned Anchor Transform (PCAT), which hierarchically reshapes anchors along both axes, and Persona-Conditioned Multi-Modal Fusion (PCMF) for BEV-level persona fusion. Training is supervised by a Hierarchical Guide Loss enforcing axis-aligned physical orderings and an Axis-Decomposed Diversity Loss preventing diagonal mode collapse. Experimental results show that PersonaDrive consistently improves over the compared baselines across multi-dimensional scenarios. The code and PCT dataset are available at https://github.com/VisualAIKHU/PersonaDrive

Abstract

Autonomous Driving Datasets

Explore similar work

PersonaDrive: Controllable Trajectory Prediction with Multi-Dimensional Driving Personas

DriveDNA: A Large-Scale Multimodal Naturalistic Driving Dataset and Benchmark for Driving Style Identification

DriveMotion: A Large-Scale Multi-Source Benchmark for Driver Motion Sequence Modeling and Forecasting