Research & projects
Six research threads follow mobility end to end, from how vehicles see to how communities move.
We benchmark how camera corruption affects camera-only, end-to-end driving, find where planning starts to break down, and test in closed loop how much degradation a policy can tolerate. For vision-language planners, we sample candidate trajectories and verify them to pick better plans.
Vehicles have limited power and strict deadlines. We plan from a few calibrated camera queries instead of a dense bird’s-eye view, run camera-based occupancy networks on FPGAs with latency that can be certified, and cut the energy of neural perception with motion-aware approximate memory.
Vehicles and infrastructure can share what they see over V2X, but bandwidth is limited. We compress shared perception into compact tokens and use cloud-assisted encoding with large multimodal models to keep collaborative perception efficient.
We regress 3D Gaussian scenes from RGB images and poses alone, localize and render large-scale scenes with mixture-of-experts coordinate encoding, and help robots explore and navigate by predicting the scene ahead.
Vehicle behavior in traffic
Tianle Zhu · Qianyi Wu · Richard Fritz
Field experiments and open datasets show how electric vehicles with one-pedal driving follow other cars, and how driver-assistance and automated driving systems interact with traffic lights and signs. We also study edge-computing operations for automated vehicles at stop-controlled intersections.
Through TRAVELS and related projects, we assess whether rural roads are ready for automated vehicles, map Georgia demonstration routes with LiDAR and HD maps, and synthesize the evidence on rural automated-vehicle deployment.
Try it · Traffic waves
Twenty-two cars on a ring road, no bottleneck. Tap the brakes,
watch the slowdown ripple backward, then add automated vehicles.
Each dot is a car’s position over time; red bands are stop-and-go waves moving backward.
Simulated at three times real speed. Human drivers follow the Intelligent Driver Model with a 0.6-second reaction delay; automated vehicles hold the average speed and keep a buffer, as in ring-road field tests.
Sugiyama et al., 2008 Stern et al., 2018
Where we work
01 · Home base
The lab’s home at the UGA College of Engineering: research, instrumented vehicles, and student projects.
02 · EV experiments
EV platoon experiments on a straight, low-traffic stretch of State Route 72, comparing one-pedal and two-pedal driving.
03 · Rural AV route · TRAVELS
A 3-mile demonstration loop past the regional hospital in Pulaski County, surveyed with LiDAR and mapped lane by lane.
Applied research
Research across vehicles,
roads, campuses, and communities.
Connect rural mobility needs with autonomous passenger-service concepts, route assessment, and demonstration planning.
UGA contributes the Georgia site work within the UW–Madison-led TRAVELS program, including local engagement and high-definition mapping.
Assess how rural road markings, signs, and digital connectivity support vehicles with driver assistance and automated driving systems.
A joint North Carolina A&T–UGA project led by Tesfamichael Getahun, with Handong Yao and Qianwen Li as investigators.
Data & demos
Browse rural-route maps, EV experiment data, and our rural AV evidence hub.
Explore data & demos
