Real-time Travel Time Estimation with Sparse Reliable Surveillance Information
summary
Claims real-time OD travel time from sparse intersection cameras by inferring missing intersection volumes and fusing static and dynamic features. Uses a modified multi-layer GCN to extrapolate traffic volumes at camera-free nodes, an encoder-decoder learned traffic pattern embedding plus static embeddings, and a multi-task learning head predicting travel time with auxiliary distance and segment-count tasks. Evaluated on three months of taxi trajectories and camera data in SIP and Shezhen, it outperforms baselines, stays stable across OD distances, and degrades gracefully as camera density drops.
AI-written summary · about
chapters (9)
- 0:00Start
- 0:17Outline
- 1:06Introduction
- 1:51Methods
- 2:16Multi-layer GCN-based Network for Real- time Traffic Statuses Inferring
- 3:16Embedding for Static Raw Features and Dynamic Road Traffic Patterns
- 3:37Multi-task Representation Learning for OD Travel Time Estimation
- 4:16Experiments
- 5:17Conclusion
related talks
46mML Use Cases at Think Big Analytics
16m(Travel) time is money! Costing turns, intersections, and ways
13mUtilizing Vehicle Trajectory Data to Characterize Performance Measures of a Highway Corridors
07mUrban Map Inference by Pervasive Vehicular Sensing Systems with Complementary Mobility
07mMIFF: Human Mobility Extractions with Cellular Signaling Data under Spatio-temporal Uncertainty
19mCycle-Level Distribution-Based Calibration of Microscopic Intersection Models Using UAV Trajectories
24mTowards Smarter Intersections Through Digital Twins - The Coupling of SUMO and SCENIMINI
16mAutomated Calibration of Traffic Demand and Traffic Lights in SUMO Using Real-World Observations