Nov 18 2020
Southwest Research Institute has received a four-year, $7,239,342 contract from the U.S. Department of Defense's Defense Advanced Re-search Project Agency (DARPA) to develop air taxi technology to test the design capabilities of artificial intelligence-augmented design systems. The project is part of DARPA's Symbiotic Design for Cyber Physical Systems program.
Air taxis are short-range electric aircraft that carry a small number of passengers relatively short distances, in comparison to commercial airliners. Many companies are currently developing different air taxi designs.
"Air taxis are a very exciting idea that also presents very a challenging engineering problem," said SwRI Senior Computer Scientist Austin Whittington, the project principal investigator. "In many ways, it's the perfect concept to test an AI's design capabilities."
Designing an air taxi presents several engineering challenges. There are a wide range of design options, from larger versions of quadrotor drones, which use four spinning rotors to lift the aircraft, to many distributed propellers for vertical take-off and landing but relying on traditional lifting surfaces (wings) for the majority of the flight profile.
Engineers led by Dr. James Walker, director of SwRI's Engineering Dynamics Department, will create pieces and parts of air taxis that AI computer design systems will use to design the completed vehicle.
"What's exciting about the air taxi is that advances in controls, batteries, and electric motors have completely opened up the design space. There are lots of potentially viable designs to be explored," says Walker.
SwRI staff, led by Whittington, will analyze the AI's air taxi designs and evaluate its design capabilities. Whittington and his colleagues will judge the designs produced by the AI-augmented design systems based on whether they meet specific criteria, such as being no larger than two cars parked beside each other, has a useful range of at least 20 to 30 miles, and is capable of carrying at least two people.
The specific requirements will be developed as part the program.
"There's less room for human error with AI-driven design," Whittington said. "AI systems are capable of thinking far outside the box and come up with concepts that people never would have."
For instance, in 2004, NASA announced that its Evolutionary software, an AI program, had designed a powerful, compact antenna for use on a collection of small satellites set to be launched into orbit around Earth. The AI software had observed millions of potential designs before delivering an unusually shaped antenna that met all of the requirements.
Whittington's team will also collaborate with Vanderbilt University researchers on modeling the air taxi's cyber-physical systems (CPS), computer systems that link sensing, control and computation between user and machine.
They are increasingly common in automobiles and can be found in nearly every commercial and military aircraft.
The two teams plan to begin work in the fall of 2020 and is expected to be completed in the fall of 2024.