Vehicle state alone predicts multiple future driving paths quickly
BLInD: Learning Driver Intent as a Distribution over Future Ego Trajectories
Computer Vision and Pattern RecognitionRobotics
Summary
Predicting where a car will go next usually depends on cameras, maps, or sensors to see the surroundings. This paper shows that just using recent driving details like speed and steering alone can reliably guess several possible future paths the vehicle might take. The authors built a fast computer program called BLInD that makes these predictions in under 3 milliseconds. They tested their approach on various driving datasets and showed it improves safety features by reducing false alarms while still detecting real dangers. This method does not need expensive sensors, making it practical for real-world cars.
What this means in practice
- •For automotive safety engineers: Use vehicle state history alone to predict multiple future driving paths and improve emergency braking decision accuracy without external sensors.
- •For embedded system developers: Deploy lightweight models running in milliseconds on automotive ECUs to enable real-time driver intent prediction for safety applications.
Authors
Flavian Pegado, Ronit Hire, Shreyas Rajesh, Soham Phade
Abstract
We present BLInD (Blind Learned Intent Distribution), a compact network that maps recent vehicle-state history (e.g. speed, curvature, indicator, and vehicle type) to a top-k distribution of future ego trajectories, with no camera, LiDAR, map, or object-track inputs. We find that vehiclestate history alone is sufficient to learn a useful multimodal distribution over near-term ego trajectories, and its low-latency nature makes it well-suited for safety-critical deployment. We investigate two distribution architectures, autoregressive (AR) and flow-matching, and train on both mixed-platform opensource and Wayve datasets. Both generalize without datasetspecific adaptation; the flow-matching model achieves best topk ADE/FDE of 0.15/0.37 m on Wayve, 0.15/0.36 m on Waymo, and 0.28/0.59 m on nuScenes, with the AR model reaching comparable coverage. Integrating the distributions into an AEB trigger task, a strict all-candidates policy reduces false positives from 1.51% to 0.11% with AR (13.7x reduction, 94.9% TP) and to 0.06% with flow-matching (25.1x reduction, 98.7% TP) compared to a 1-CTRV policy with 100% true positive score. BLInD runs in 0.87 ms with the AR head and 2.9 ms with the flow-matching head on an NVIDIA DRIVE Orin ECU making it compatible with real-time deployment on automotive ECUs. While existing learned distribution models rely on scene context and blind vehicle-state models typically collapse to a single path, BLInD is learned, blind, and cross-domain simultaneously, a combination not demonstrated by prior work. These results show that such a distribution provides a controllable and plausible intent sampling interface for downstream systems, with AEB as one instantiation.