Future Innovator spotlight / Akujobi building device to cut construction risks

“Eyes in the back of your head.” For moms, absolutely. For construction workers, not so much.

But there could be a time when construction workers have them, too. John Akujobi, a senior computer science major, is working on a project to do just that. He is in his second year of an effort to create a wearable device that would alert construction workers to a hazard, such as a sudden drop-off, such as the edge of a roof, or an approaching vehicle.

He is working on the project under the supervision of assistant professor Chulwoo Pack as a Future Innovator of America.

Akujobi is one of eight recipients in the fourth class of Future Innovators. They are selected by the Jerome J. Lohr College of Engineering and are awarded $5,000 with $4,500 as a stipend and $500 to cover the cost of lab supplies or travel to disseminate the results of their projects.

The fellowships were created to provide unique research opportunities for undergraduate students in the college. Any student is eligible to apply as long as they are attending full time and have a GPA of 3.0 or higher.

Each student works with a potential project mentor, who must be a faculty or research staff member, to develop and submit a research plan that entails learning by doing.

Project began in January 2025

Akujobi submitted his original proposal in December 2024 and began work on it in January 2025. By mid-March 2025, the hardware testbed of the prototype was virtually complete. Since then, he has been working on refinements to improve the device’s accuracy and redundancy as well as making the construction hardhat harness easier to wear.

He has named his project AMBER, short for “Affordable Multimodal sensor-Based Environmental Risk detector.”

“The success of this project will impact multiple stakeholders by giving workers timely alerts about hidden hazards, helping companies reduce accidents and improve productivity at a lower cost, and offering regulators a practical tool to strengthen safety standards,” Pack said.

Safety at a reasonable price

While the construction industry has tools for this, such as harnesses, cables and railings, most kick in during or after a fall, not before one. Fixed barriers have another problem — they can't follow the worker. A roofer on a pitched surface has nowhere to anchor one.

“AMBER isn't meant to replace any of that. It's an add-on, a layer of awareness for the places and moments where barriers can't go,” Akujobi said.

“Many solutions rely on high-end sensing devices and large-scale artificial intelligence models, which come at a cost that hinders their practical adoption for individual worker safety,” Pack said.

Akujobi, Pack and Phuong Nguyen, an assistant professor in mathematics, reviewed the existing research and found one company building something similar. It uses multiple hubs across the job site, an enterprise system monitored through cameras on the helmets.

"This makes it very expensive, limiting it to large construction companies, and the head-worn cameras raise privacy concerns for the worker," Akujobi said. "AMBER gives independent workers access to this kind of safety technology while respecting their privacy and putting them in full control of their data."

Working in his off-campus apartment and earlier in a lab in the basement of Daktronics Engineering Hall, Akujobi has now produced twin prototypes, alpha and beta, with the first allowing for rapid testing of new ideas.

He estimates the material cost of the module at under $100, a fraction of what comparable industrial safety systems cost, which can exceed $1,000 per smart-helmet unit. AMBER also attaches to any existing construction helmet.

Sensors — each with a different duty

AMBER features nine sensors across three types: ultrasonic, time-of-flight and TF-Luna LiDAR. The TF-Luna is a narrow-beam sensor offering range detection from 0.2 to 8 meters; the time-of-flight sensors detect from 0.04 to 4 meters.

The sensors cover four distinct directions; downward to the left, downward to the right, downward and directly behind, and straight out behind the wearer.

Using three sensor types rather than one is a deliberate design choice. Environmental conditions affect each differently — “very bright lights could clash with the time-of-flight sensor, but the ultrasonic sensor would still be working well,” Akujobi said. “By pairing them together, we increase accuracy and redundancy.”

The brains of the harness is a Raspberry Pi Zero 2W, a compact computer that gives Akujobi the flexibility to iterate quickly on software. In the version last year he tested a Raspberry Pi Pico 2W, Teensy 4.1 and ESP32 before settling on the current platform. While still bulky, the harness has also become lighter with heavy metal mounting pieces replaced by small screws and an interlocking plastic grid made by 3M.

Building the software to run on it has been a major undertaking. Akujobi has written more than 7,100 lines of Python code across 30 modules — creating classes for each sensor type, building the fusion algorithms that combine their readings, and writing the classifiers that decide when a hazard is real. The codebase is backed by 550 automated tests.

In February, Akujobi collected initial live field data with the ultrasonic sensor suite by navigating through Daktronics Engineering Hall and climbing tables. He hopes to have all sensors and alert systems tested and more field data collected by the time he graduates in May.

AMBER work to continue in grad school

Ultimately, Akujobi would like to replace the full computer in the harness with a microcontroller for faster processing, reduce the weight by switching from prototyping cables and boards to a printed circuit board, and add local AI options to allow for better on-device predictions.

“Ongoing research is to see if we can use edge detection with computer vision to improve the performance while keeping the all data and processing fully on-device without any need for external systems," Akujobi said.

The alert system uses four hazard levels; low, medium, high and critical. At every level, all three modalities engage together: vibration, directional LEDs to identify which side the hazard is on and a buzzer provides audio warning. What changes is intensity. At low, slow gentle pulses serve as an awareness signal; at critical, all devices fire continuously at full power.

Akujobi plans to continue his studies at SDSU as a graduate student, giving him more time to advance AMBER. Eventually, he would like to start a company and build simpler versions of AMBER, such as a single sensor that keeps the wearer informed to what is behind them. Such product could improve situational awareness for warehouse workers, law enforcement, farmers, firefighters and those working around moving heavy machinery.

Pack said, “What stands out most about John is the way he keeps pushing the project forward. Every week, he came in with new trials, new findings and new ideas. He was always eager to explore different hardware options to improve the system. His attitude and work ethic are truly outstanding."

Tragic accident spurs AMBER project

The impetus for AMBER was an accident in Akujobi's home country of Nigeria. He recalls a construction accident in which a bricklayer backing up a wheelbarrow didn’t realize his proximity to the edge of a four-story scaffold and fell to his death. The incident stuck with him.

In the U.S., 150,000 workers are injured annually from preventable worksite accidents with falls being one of the leading causes of death.

Through computer science studies and conversations with his friends and faculty at SDSU, Akujobi discovered the power of sensors, algorithms and machine learning. He realized those things hold the potential for preventing such future tragedies.

“AMBER doesn’t prevent you from falling, but it gives you a sixth sense,” Akujobi said.

AMBER is only one of the irons Akujobi has in the fire. He's also on a senior design team with two electrical engineering students and two computer science peers building Impact Cap, an electronic helmet for children who self-injure that alerts caregivers and detects environmental sensory triggers.

He also works a couple of data-driven jobs on campus. He does website design, data entry, analytics and tabulation for SDSU’s Division of Research and Economic Development and creates software for managing data assets for KBR, a contractor with the U.S. Geological Survey, which runs EROS Data Center near Baltic.

Akujobi will share the story of AMBER at SDSU’s Student Research, Scholarship and Creative Activity Day April 14.