WatchDog: Real-time Vehicle Tracking on Geo-distributed Edge Nodes

Vehicle tracking, a core application to smart city video analytics, is becoming more widely deployed than ever before thanks the increasing number of traffic cameras and recent advances in computer vision machine-learning. Due constraints bandwidth, latency, privacy concerns, tracking tasks are preferable run on edge devices sitting close cameras. However, provisioned with fixed amount computing budget, making them incompetent adapt time-varying imbalanced workloads caused by dynamics. In coping this challenge, we propose WatchDog, real-time vehicle system that fully utilizes nodes across road network. WatchDog leverages different resource-accuracy tradeoffs, decomposes schedules judiciously based current workload maximize while ensuring provable response time-bound at each device. Extensive evaluations have been conducted using real-world city-wide trajectory datasets, achieving exceptional performance guarantee.