Projects
Systems that Operate 24/7/365
A selection of work across transit intelligence, UAV vision systems, and control systems.
Large-Scale Transit Signal Priority
Designed two large-scale, cloud-based TSP systems coordinating thousands of buses and traffic signals. Real-time GPS reports are processed in the cloud, intersection arrivals are predicted using AI models, and TSP commands are sent to the traffic signal controllers.
Traffic Signal Green Band Prediction
A live 1/sec traffic signal feed is trained to predict the upcoming green band. Real-time 1/sec GPS bus data is used to predict intersection arrival times. An onboard display device instructs operators to dwell or slow to roll through on green.
Fleet Orchestration at Scale
Designing a system to track, orchestrate, and control thousands of buses and tens of thousands of operators for deployment at major sporting events such as the World Cup, NFL, concerts, and more.
Large-Scale Transit & Traffic Delay Analysis
Large-scale analysis of transit and traffic delay at traffic signals, using both high-frequency GPS and GTFS-RT feeds to identify transit “chokepoints.”
Control Systems — Gate Control
A fully fault-tolerant railroad gate control system operated from the cloud to the wayside, employing multiple cellular routers, cell providers, and intelligent failover.
Commercial Delivery UAV — Auto-Landing
One of two UAV vision systems we developed: a commercial delivery drone vision system trained to land autonomously at a predefined target. Deployed on a very low-cost COTS platform.
Military UAV — Vehicle Tracking
A military application using NVIDIA GPUs designed for real-time detection and tracking of specific vehicle types. Object detection was combined with optical flow. Also developed a web-based semi-automated image training system.