Research on MOTIF Lab

Traffic flow theory

Mon, 01 Jan 0001 00:00:00 +0000

Traffic flow theory remains the structural language of transportation engineering. Our work in this thread strengthens that language for an era in which vehicles are no longer interchangeable units. We study car-following and lane-changing behavior, the conditions under which fundamental diagrams hold or break, and the propagation of disturbances in heterogeneous traffic streams.

Recent questions include: How does the variance in driver behavior — amplified or dampened by ADAS — change capacity at active bottlenecks? When mixed-autonomy traffic violates the assumptions of LWR-class models, what minimal extensions recover predictive power without overfitting? When does signal-control theory remain robust in congested urban networks, and when does it not?

Vehicle technologies and traffic impact

Mon, 01 Jan 0001 00:00:00 +0000

A vehicle is no longer a fixed object in a flow model — it is a moving system whose perception, control, and powertrain shape its behavior. We characterize ADAS-equipped and electric vehicles empirically and analytically, then quantify what their adoption implies for capacity, stability, and safety on existing roadways.

Specific work includes: car-following calibration for production ACC and LKA systems; the energy-efficiency / throughput trade-off introduced by EV speed-power profiles; and how partial market penetration of automated features changes the distribution of headways and the probability of rear-end conflicts. We collaborate with vehicle dynamics researchers and rely on a mix of test-track data, naturalistic driving data, and microsimulation.

AI for mobility

Mon, 01 Jan 0001 00:00:00 +0000

Modern vision and learning systems extend the reach of traditional traffic monitoring. We develop and evaluate AI tools that turn dash-cam video, roadside cameras, and probe data into operational signals — vehicle counts, incident alerts, pavement and roadside hazard detections — at a fraction of the cost of fixed instrumentation.

Our emphasis is not on novelty in the model architecture alone but on closing the loop with traffic operations: what false-positive rate is tolerable for an alert that triggers a maintenance dispatch? How do edge-device and bandwidth constraints shape what is actually deployable? How do we evaluate detection performance against the messy ground truth available from agencies?