- UF won an Air Force Research Laboratory grant for autonomous systems research.
- It will establish the center for Designing Integrated, Generic and Interoperable Compositional Analyses of Modern Systems.
- The project is designed to improve coordination and performance of drone swarms and complex systems.
Imagine a cluster of drones swarming above a dense thicket of trees, eyeing its way through unfamiliar topography to locate a missing person amid unpredictable weather.
Each drone has its own complex communication and computing capabilities. But as that weather starts to sour against the clock, the rescue mission requires group optimization, not a single drone with its own problem to solve.
The team driving the drones needs to act quickly to determine how many drones to deploy, how they will communicate, which trajectory to take and how they will adapt to challenges such as wind. To be most effective, that swarm must formulate a problem that accounts for flight constraints, solve that problem to determine the route and then rely on a controller to stay on the same path.
A University of Florida engineering research team is working to optimize such communications among autonomous units — drones and self-driving vehicles, for example — in a project funded by an Air Force Research Lab University Center of Excellence Award.
“The proposed research plan will develop compositional methods for decision-making in optimization, dynamical systems and control, which will produce new, higher-performing decision procedures than the current state of the art in each area,” noted project lead and UF engineering Assistant Professor James Fairbanks, Ph.D., in the project’s summary.
The $3.5 million AFRL grant will establish the center for Designing Integrated, Generic and Interoperable Compositional Analyses of Modern Systems (DIGICAMS), which will develop unified mathematical frameworks for the decision-making processes across optimization, dynamical systems and control.
The center brings together a team of researchers with deep mathematical expertise and a passion for solving real-world problems.
The research plan centers on developing mathematical structures across multiple disciplines. The team plans to develop foundational mathematical formalisms for unifying tasks, systems, decision procedures, constraints and scales of a wide range of engineering systems.
Ultimately, the project will yield large-scale computational tools for the U.S. Air Force that specify and solve complex engineering problems with autonomous systems.
“We will develop such a platform and use it to distribute the methods to Air Force researchers and bench engineers,” noted Fairbanks, who works with the Department of Mechanical & Aerospace Engineering.
While much of the research will be at the chalkboard and behind computers, there will be field work at UF’s Autonomy Park, a facility where researchers investigate heterogeneous networks of multi-agent autonomous systems operating in contested environments.
Why is the U.S. Air Force’s interested?
Optimization of complex systems.
“For example, when you want the optimal performance of an aerospace system, you need optimal performance from all your components such as propulsion, control, guidance, and you need to find a way to combine them that works together,” Fairbanks said.
Optimization, he added, is critical to developing engineered systems in every domain. Developing systems in operations and logistics, electronics, mechanical and aerospace all rely on optimization to make decisions. The U.S. Department of Defense calls it “systems of systems.”
“On aircraft carriers, for example, there are many systems aboard each carrier. Each ship has multiple systems, but then the carrier group is a system of systems,” he said. “Each ship is a system of its components, and then the system of ships is a system of systems,” Fairbanks noted.
In the above drone-rescue scenario, the key engineering considerations are optimization, dynamics and control. Engineers typically analyze each factor individually with different mathematical approaches.
“While each individual subsystem may be optimal in isolation,” the researchers noted in the research summary, “their aggregated composite may be suboptimal in ways that are difficult to detect. This lack of system-level understanding can lead to overall instability, even when using stable subsystems, or cause a persistent suboptimal allocation of resources, and more.”
Decisions are deeply interconnected.
Adding more drones could improve coverage but also strain communication, leading to instability. Solving such problems, the team contends, requires a holistic mathematical framework that connects these layers. This, Fairbanks said, is exactly what DIGICAMS will develop using category theory.
The team also includes College of Engineering Interim Dean Warren Dixon, Ph.D., Matthew Hale, Ph.D., at Georgia Tech and Miroslav Payek, Ph.D., at Duke University. The non-academic partner is the Topos Institute in California, which develops and applies category theory to real-world problems.