Story by Wynne Parry  |  lllustrations by Matthieu Forichon
 

By combining robotics, automation, and artificial intelligence (AI), Maryland Engineering researchers and students are building solutions that can help save lives, protect property, and safeguard the environment. 

One such system, RoboScout, Maryland’s entry in a competition run by the defense agency DARPA, would rapidly survey and assess injuries at sites of disasters or violence. The information RoboScout is designed to collect could help first responders prioritize medical care, a process called triage. The ambitious project, while in its early stages, is an innovative step toward autonomous systems to help as many people as possible survive such events.

Here’s how RoboScout would work.

 

Step 1: Arrival

At the site of disaster or violence, medical personnel set up the RoboScout system, which consists of drones custom-built by Maryland engineers, robot ‘dogs’, and a base station that runs sophisticated AI models.

Step 2: Initial Survey

The operator launches drones equipped with visible light and infrared (or thermal) cameras. The location of each person they identify appears on a map on the base station’s monitor.

Step 3: A Closer Look

The drones hover closer to those injured. Their sensors collect essential health data on bleeding, traumatic injury, and difficulty breathing, and send it to one of the base station’s AI models called an inference engine.

Step 4: On the Ground

The robot dogs are dispatched to those injured to collect more information using their cameras and radar sensors. An AI called a large language model that is located on the dogs can talk to the injured person through the dogs’ two-way radio.

Step 5: Report

The inference engine weighs observations from the drones and dogs to assess each person’s injuries, like a first responder would. “We’re trying to recapture the performance of a human medic assessing injuries, and scale it to situations that are dangerous or where there aren’t enough medics to look at everybody in the first 10 minutes,” says Derek Paley, director of the Maryland Robotics Center and RoboScout’s team leader.

Scalable, Timely & Accurate Triage

 

In mass casualty incidents, whether civilian or military, triage needs often outstrip resources and place enormous strains on medical personnel. But what if we could ease this burden by deploying advanced technology? In the DARPA Triage Challenge, participants are developing novel methods of detecting injuries so that medical personnel can respond more quickly, efficiently, and precisely.

$750k
Per Year Funding Received by RoboScout & Other Funded Teams
3rd
RoboScout's ranking, out of 11 teams, in Year 1 of the competition in October 2024
October
Year 2 competition occurs this fall, and the final competition will be in 2026

“A robotic autonomous system has three parts to it: perception, planning, and action,” says RoboScout co-team lead and Distinguished University Professor Dinesh Manocha. “It's similar to a human in that way.”

First responders must work quickly, but with little information to go on, says Dr. Sarah Murthi, co-lead of the RoboScout team and professor at the University of Maryland School of Medicine. “They don't have a sure way of knowing, for example, if the patient to the left is much sicker than the patient to the right, and whether the patient to the left should be extracted first.” With a system like RoboScout, she says, “you could start getting much more precise data about each patient within minutes.” Considering ethical issues—the AI’s priorities, for example—will be important, says Murthi. “How do we ensure its decision making is intended to save the most lives?”

Autonomous rescue systems could:

  • Move the injured to safety, allowing them to receive medical attention more quickly and protecting first responders.
  • Make complex decisions, based on data, on injuries and medical resources. 
  • Attach sensors to the injured to continuously monitor their condition.

 

 

Assisting first responders is just one example of the potential good autonomous tech can do. UMD engineers are building systems that address a multitude of challenges, while also preparing current—and future—workers for this innovation revolution.

 

 

Monitoring Wildfires

 

Globally, the severity of wildfires is increasing—they spread faster and burn larger areas at higher intensities. Maryland engineers are developing new, innovative tools to keep up with the increasing threats and mounting challenges posed by extreme wildfire events. 

Wildfires can ignite in remote places. But all too often, their destructive fury encroaches on homes and communities. In January 2025 alone, wildfires in California burned more than 57,000 acres, costing 29 lives and destroying more than 16,000 homes and other structures. Maryland engineers are designing drone-based systems for autonomously assessing, and even fighting, fires like these from above.

Two teams are working on separate projects with related goals. The first, funded by a UMD Grand Challenges Grant, aims to develop drones that fire departments could fly over fires to collect information about them in real time, so they can make more informed decisions about how to fight them. The machine’s AI would possess the capacity to navigate along a fire on its own. All the while, it would assess images from its visible light and infrared cameras to identify the burning areas below. This data would be transmitted to the ground, where firefighters could use it to find a fire or modify their strategies for fighting it. 

A second team, called Crossfire, is participating in the XPRIZE Wildfire competition, which challenges participants to develop autonomous technology capable of detecting an early-stage wildfire somewhere in a 1,000 square-kilometer area, and then suppressing it (all within 10 minutes). To put out a fire once it’s identified, the team is experimenting with a well-established fire suppressant: water. This task would likely be carried out by a different drone or drones, which would release a water-carrying vessel that ruptures at a predetermined height above the fire.

Ultimately, Fernando Raffan-Montoya, an assistant professor of fire protection engineering leading the first wildfire project and collaborating on Crossfire, envisions systems like these informing on-the-ground decisions, such as when to order evacuations. Drone-based fire suppression, meanwhile, could one day protect communities from an approaching wildfire, he says.

Between climate change, which creates conditions more conducive to fires, and increased building in wildland areas, the problem will continue to grow. “We believe that autonomy will bring a much-needed part of the solution,” Raffan-Montoya says.
 

 

Student Spotlight: Help Where Humans Can’t

A drone’s cameras can capture a small fire all at once. But, more often, fires are too big for a single image. Andrés Felipe Rivas Bolivar, a Ph.D. student in aerospace engineering working on the Grand Challenges wildfire project, is developing an image recognition algorithm to identify the edge of a fire. With this information, the algorithm generates a trajectory along which the drone navigates to capture the fire’s full scope.

Raffan-Montoya recruited Rivas Bolivar from Colombia, where Rivas Bolivar had designed and built drones first as a hobby and then through his master’s program in mechanical engineering. Rivas Bolivar envisions a drone-based monitoring system being used in places, including Colombia, where fire departments may not have the resources to monitor fires with manned aircraft.  

“We’re just in the investigation phase now,” Rivas Bolivar says. “But at the end, if we have a platform that does all the things we are thinking of, it could help firefighters save money and lives.”
 

 

 

Urgent Medical Delivery

Maryland’s Smith Island doesn’t have a pharmacy, so to pick up a prescription, any one of the roughly 200 islanders must take a boat across the Chesapeake Bay to the mainland. But options are limited since ferries may run only once a day, or less frequently in bad weather. 

Supported by a $1.76M U.S. DOT grant, a pilot program utilizing drone technology could accelerate medical shipments to and from the historic fishing community of Smith Island, which straddles the Maryland–Virginia border in the Chesapeake Bay, 10 miles west of the nearest mainland. (Photo courtesy University of Maryland UAS Research and Operations Center)

These logistics can be particularly problematic for residents with chronic health conditions—diabetes, heart disease, and the like—who must take medication regularly. What’s more, urgent situations occasionally arise when someone needs a prescription on short notice. 

“There are cases when 24 hours can make a big difference,” says John Slaughter, director of UMD’s UAS Research and Operations Center located in southern Maryland. 

Working with partners, including the state of Maryland, Slaughter leads a team that is developing a solution: drones that deliver medications to the island and perhaps bring back samples, such as vials of blood, for testing. With this autonomous system, they aim to help improve the health of island residents, particularly the most vulnerable among them who may have ongoing medical needs and trouble traveling.

The team hasn’t settled on a drone model yet, but because the flight is relatively long, they plan to use an energy-efficient machine that takes off and lands like a helicopter, but flies like an airplane between locations. Once they have selected a drone they intend to modify it to handle a less-than five pound (0.45 kilogram) payload that includes the camera it uses for navigation. 

They intend to give the drone’s AI what’s called “conditional autonomy,” Slaughter says. “It’s going to be able to get itself to the delivery location, with the pilot ‘riding along,’ as it were, to monitor its progress and intervene if there’s a problem.”

 

More Productive, Sustainable Oyster Farming


Supported by a $10M USDA grant, Maryland engineers are developing novel technologies and a sustainable management framework to help farmers tap the economic potential and environmental benefits of shellfish aquaculture, which until now has been bottlenecked by outdated tools and methods. (Photo by John T. Consoli)

When scattering young oysters to seed a new crop, the Chesapeake Bay’s oyster farmers have no way of knowing for certain that the seafloor provides the right conditions for the bivalves to grow. Likewise, when dragging a dredge to retrieve their crop, a farmer has limited information about where best to retrieve mature oysters, leading to unnecessary collateral damage to the ecosystem below.

A team led by Miao Yu, a professor of mechanical engineering, intends to update these long-standing strategies to make farming more efficient and productive, and less destructive. “We want to change the current practice from essentially a random approach to precision planting and harvesting,” she says. 

Yu is leading a $10 million USDA-funded project that aims to make shellfish farming smarter and more sustainable. As part of it, she and her colleagues are developing robotics and AI technologies that will autonomously survey the seafloor of shellfish farming plots like those in the bay. 

A submerged robot with a camera would collect detailed images underwater while tethered to a boat or a robot on the water’s surface. This second robot would be equipped with a sonar device that emits pulses of sound to create an acoustic image over a larger field of view. Based on this information, Al algorithms would create a map of the farm and assess the crop, including, for example, the density of the oysters at a given location. A separate AI model would generate recommendations for farmers, including the GPS coordinates for a path to follow when harvesting.  
  
Ultimately, the team aims to help the farmers improve their oyster yields by at least 10 percent. This more strategic approach could offer other benefits too, including protecting the seafloor and reducing fuel consumption. Increasing production of oysters could also provide a source of protein responsible for lower levels of greenhouse gas emissions than other sources of this nutrient, notably beef.

 

 

Training for the Revolution

With their ability to function independently, autonomous systems can partner with human workers and take on the least desirable, most dangerous jobs. UMD’s MATRIX Lab aims to ease this transition by training workers, and the students who will one day fill these roles, to adapt to this technology future.

At the southern anchor of Maryland’s autonomy corridor in the MATRIX Lab, students can access unparalleled resources for innovating the next generation of AI and autonomous systems. (Photo courtesy of TEDCO)

Located in the Southern Maryland Autonomous Research and Technology (SMART) Building, the MATRIX Lab offers programs that run the full educational gamut, training students at the high school, undergraduate, and graduate level, as well as those already in the workforce. Industry-focused short courses help participants acquire new technical skills that can aid them in their current positions and help them advance in their careers. In a recent course, participants learned about how drones are designed and built, how to operate them, and the Federal Aviation Administration’s requirements to certify them.

While topics like this are relevant to autonomy, MATRIX has begun developing a program focused solely on it: a new Master’s of Engineering in Test and Evaluation. This program—the first of its kind in the U.S.—will focus on assessing how autonomous systems perform.

“Nobody has figured out how to do that yet, because autonomous systems make their own decisions,” says Matt Scassero, MATRIX’s director of operations and outreach. “MATRIX is developing an approach tailored to this challenge, working with its partners in government, industry, and academia.”

Autonomous systems will inevitably reorganize the workforce, Scassero says. He emphasizes that, in the end, everyone benefits: “Just like the steam engine, the circuit board, and the Internet, AI and autonomous systems will open up new areas of productivity that we didn’t even imagine before.”