Beyond Direct Sensing: Harnessing Indirect Observations from Third-Party Sensors in Vehicle Tracking
Gaofeng Dong*, Vamsi Eyunni*, Pragya Sharma, Kang Yang, Mani Srivastava
University of California, Los Angeles · *Equal contribution
Project page · Paper (arXiv, coming soon) · Citation
GrayTrack combines sparse direct observations with weak, anonymous passage events from third-party sensors using a road-constrained particle filter. A CARLA–Mininet-WiFi testbed infers these events from encrypted camera-traffic metadata.
In controlled CARLA Town05 experiments, indirect observations reduce Road-PF trajectory RMSE from 92.3 m to 36.8 m (60.1%) and catastrophic track loss from 35.8% to 0.3%. Passage detection achieves 0.989 F1 on 242 held-out camera sequences.
| Folder | Contents |
|---|---|
| Part1-testbed | CARLA capture, Mininet-WiFi replay, packet grouping, and passage detection. |
| Part2-RoadPF | Road-constrained particle filtering, tracking baselines, and evaluation. |
| docs | GitHub project page. |
Each part has its own setup instructions and dependencies.