The authors propose TriDrive, a framework that encodes driver kinematics (via an anchored representation with BATON pose pretraining), vehicle dynamics (from CAN-bus data), and road context (using frozen V-JEPA 2 latents combined with structured road margins).
Unified model jointly forecasts driver, vehicle, and road states for monitoring
TriDrive achieves state-of-the-art kinematic forecasting on AIDE and improves warning timeliness in an on-road study with 14 drivers.
Big Tech
Yuhang Wang · Jingxin Yang · Chuheng Wei · Yuechen Guo · Jinghan Xu · Zhao Han · +1 more
University of South Florida · NVIDIA · Purdue University · Hunan University
Research Digest··3 min read
The authors introduce TriDrive, a unified framework that jointly models driver kinematics, vehicle dynamics, and road demands using modality-specific encoders and directed residual connections.
Why this paper
From NVIDIA and 3 others · Released weights
In one line
TriDrive jointly forecasts driver kinematics, vehicle dynamics, and road demands, setting a new state of the art on AIDE and improving driver monitoring on BATON.
What it released
Weights
What we could check
- ·No code link found
- ✓Weights released (huggingface.co)
- ·No dataset link found
- ✓Compute or model size stated (hardware comma four with external 8 GB GPU)
- ✓Limitations stated by the authors (4 noted)
- ✓Reports numbers on named benchmarks (5 benchmarks)
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§