Researchers Develop Drone-Based Wireless Sensor Network for Underground Reconnaissance
09/01/2023
By Ed Brennen
Disorienting darkness. Claustrophobic confines. Attacks possible at every turn—with no way to call for help.
Every battlefield is dangerous, but perhaps none more so than an underground tunnel behind enemy lines.
To help soldiers safely scout subterranean passages, researchers from the Francis College of Engineering are developing autonomous drones that can fly in advance of the soldiers and detect signs of danger—like a modern-day canary in the coal mine.
“UMass Lowell has the expertise for remote sensing in the army has the need, so we are working together to develop the technology to ensure their personnel are safe.” -Electrical and Computer Engineering Prof. Xuejun
“UMass Lowell has the expertise for remote sensing and the Army has the need, so we are working together to develop the technology to ensure their personnel are safe,” says Electrical and Computer Engineering Prof. Xuejun Lu, principal investigator on the project, which is supported by a $956,400 grant from UML’s Harnessing Emerging Research Opportunities to Empower Soldiers (HEROES) collaborative with the Army.
Since 2020, Lu and his two co-PIs—Electrical and Computer Engineering Prof. Chunxiao (Tricia) Chigan and Civil and Environmental Engineering Prof. Yuanchang Xie—have been integrating advanced sensors, infrared cameras and communication devices with drones to show how they could be used by the military for underground reconnaissance missions. The project, a collaboration with drone maker Asylon Robotics, could also have civilian applications, such as monitoring hazardous chemical spills or looking for earthquake survivors in confined spaces deemed too risky for first responders.
Several students and postdocs are involved with the project, including Mohamed Martini ’17, ’20, ’22 (computer engineering), Boyang Xiang ’17, ’22 (electrical engineering), Qu Liu ’20 (computer science) and Yan Cui ’19 (computer engineering).
Two of the project’s biggest challenges, Lu says, are making sure that the sensors are small enough to fit on a drone and that the laser beams they emit to detect gases do not require too much power.
That is true, as well, for the work Xie is doing with the drone’s thermal cameras, which are used to detect signs of human activity in tunnels. An onboard computer uses machine learning algorithms to quickly identify objects that the enemy may have left behind, such as soda cans, wine bottles and cigarettes.
“If we detect something suspicious, process all the data on the drone and send only relevant data back to the command center, we can save on communication cost and also power,” Xie says.
To make this possible, Xie and Martini recorded various objects with thermal cameras, annotated the images and then fed the data into deep-learning neural network models.
“Once the model is trained, you don’t need to teach it again; it can just do the detection automatically,” says Xie, who has implemented similar algorithms on a research project that uses drones to study traffic patterns on highway ramps. “The technology itself is not new, but this is an interesting application.”
Chigan, whose background is in wireless communication and network security, is responsible for making sure that the drones’ data transmission is reliable, power-efficient and secure.
Because the range of radio signals is limited in an underground environment, Chigan uses multiple drones, or nodes, to form a “mesh wireless sensor network.” The data collected from the drones’ sensors can then be transmitted to a gateway node, which could be positioned at the tunnel entry.
Continuous user authentication—making sure that the enemy is not intercepting the drones’ data with fake nodes during the mission—is the most important security feature of the project, Chigan says.
After building the drones and then testing them at the North Campus parking garage, the research team demonstrated the object-detection capabilities of the drones last fall in a darkened Moloney Hall at University Crossing. They are now working to fuse data from the drones’ infrared cameras with their existing lidar (light detection and ranging) sensors so they will have the capability to create detailed 3D maps of their target areas.
Because their research work is typically more theoretical, Lu, Xie and Chigan agree that building a fully operational drone—which can ultimately be used to help protect soldiers—is especially rewarding.
“As professors, we publish papers and do mathematical analysis and computer simulations, but we don’t always have a testbed to see if what we study can really be implemented,” Chigan says. “It’s exciting to see the real application in action.”