Engineering Safer Streets: An Interview with Muhammad Fahad on AI, Computer Vision, and Smarter Crosswalk Lighting

Muhammad Fahad is a research assistant at the University of Wisconsin–Milwaukee whose work sits at the intersection of computer vision, trajectory modeling, and robotics. One of his most visible initiatives is the Enhanced Crosswalk Lighting Evaluation Project, a $125,000 effort funded by the Wisconsin Department of Transportation (WisDOT). Muhammad Fahad authored the winning proposal and now leads the experimental deployment, real-world testing, and data analysis of a system that helps cities quantify the impact of lighting on pedestrian safety.

In this interview, Muhammad Fahad explains his path, the technical backbone of his method, and how the work is being used beyond campus to help engineers, policymakers, and communities make safer decisions.

1) Please tell us more about yourself.

My name is Muhammad Fahad, and I’m a graduate research scientist at the University of Wisconsin–Milwaukee, specializing in AI, computer vision, and intelligent transportation systems. Over the past several years, I’ve focused on practical technologies that enhance roadway safety, particularly through the use of panoramic video, trajectory modeling, and GNSS-enhanced evaluation systems. I am the lead researcher and testing architect for the WisDOT-funded Enhanced Crosswalk Lighting Project, where I designed a first-of-its-kind framework that labels pedestrian-vehicle interactions based on real-world risk indicators. I authored the winning proposal for this $125,000 award, and I now lead data collection, experimental design, and system implementation. Beyond the technical work, I am committed to developing tools that cities can utilize to enhance pedestrian safety and inform more informed public investments.

2) Muhammad Fahad, how did this project start?

This began with a simple gap: cities were investing in crosswalk lighting but lacked a scientific way to measure whether it improved driver behavior or pedestrian safety. I proposed a new approach using panoramic cameras, RTK-GNSS, and AI-based trajectory analysis to build a maneuver-level safety evaluation system. After WisDOT awarded the proposal, I took on leadership of the entire data collection and analysis framework. This is not a lab-only exercise; it’s deployed at real locations in Milwaukee and Madison and was built from the ground up to bridge engineering and public impact.

3) What’s technically original about your method?

The method goes beyond static lighting metrics and evaluates actual driver and pedestrian behavior. We capture crossings with a panoramic video system, detect and track road users with YOLOv9e and DeepSORT, project trajectories to the road plane via homography, and validate positions with high-precision RTK-GNSS. The distinctive step is applying behavioral and physics-based thresholds critical gap acceptance, stopping-sight-distance, and braking profiles to each trajectory. That lets us classify interactions as safe, marginal, or high-risk based on quantifiable indicators. Agencies can then compare lighting configurations and know, with data, which option improves safety.

4) What role are you playing day to day?

I lead the technical program end to end system installation, calibration, data fusion, analytics, and reporting. I design the test protocols, configure sensors, run evaluations, and oversee the labeling pipeline. I also coordinate with city engineers, WisDOT partners, and lighting manufacturers. This isn’t just a paper it’s a live system producing evidence decision-makers can act on. Every threshold, algorithm, and output has been developed or implemented by me to ensure accuracy and reproducibility.

5) What results are you seeing so far?

Our initial deployment in Milwaukee captured more than 180 vehicle-pedestrian interactions. The evaluation system achieved 86.76% accuracy, 78.08% precision, and 72.84% recall in detecting and labeling high-risk maneuvers strong performance for an uncontrolled, real-world setting. We are now testing different lighting configurations mounting heights, color temperatures, and glare controls and we’re already seeing measurable improvements in recognition and early braking. These findings are feeding into a technical specification matrix agencies can use when selecting lighting and verifying performance before installation.

6) Why does this matter beyond Wisconsin?

Nighttime pedestrian risk is a national problem. Many serious injuries occur in low-light environments, yet most agencies lack a reliable, cost-effective way to test lighting effectiveness before spending on upgrades. Because our approach uses commercially available sensors and a replicable pipeline, it scales to other cities and can also support related applications such as AV perception and intersection design.

7) How have partners responded?

Feedback has been very positive. WisDOT staff have called the work a breakthrough in translating lighting design into measurable safety performance. A city engineer told me, “We finally have a way to justify lighting upgrades with evidence, not just hunches.” Manufacturers have asked to use the protocol for product validation. That kind of external interest signals real-world value and accelerates adoption.

8) What comes next for the Enhanced Crosswalk Lighting Project?

We’re expanding to additional sites, including winter-weather trials in Madison to evaluate performance under snow and fog. I’m integrating the framework with my 1/10-scale autonomous vehicle platform to study how lighting influences machine perception as well as human drivers. I’m also preparing a lighting diagnostic toolkit calibration guides, ground-truth alignment utilities, and an open-source risk-labeling script so other communities can adopt the process without a large research team.

9) How does this fit your long-term vision?

My aim is evidence-based, low-cost infrastructure that can be deployed anywhere from a downtown corridor to a rural town with no traffic lights. With a background spanning autonomous vehicles, AI, and robotics, I build systems that don’t just observe the world but actively make it safer. Engineering should solve public problems; this project reflects that philosophy and sets the stage for broader work on human- and machine-vision-informed safety.

10) What advice would you share with young engineers?

Lead something that matters. Write the proposal, build the tool, and measure the impact. Document your role, funding, and outcomes, but never lose sight of the people your system serves. When technology measurably improves everyday life, everything else follows.

Project Overview

Project Title: Enhanced Crosswalk Lighting Evaluation
Funding: $125,000 Wisconsin Department of Transportation
Institution: University of Wisconsin–Milwaukee, College of Engineering & Applied Science
Principal Investigator: Prof. Xiaowei (Tom) Shi
Project Lead, Testing, Evaluation & Analysis: Muhammad Fahad
Impact: Data-driven lighting guidance for municipalities; scalable, field-ready tools for pedestrian safety

Media Contact
Muhammad Fahad
Website: https://www.fahadmuhammad.com/