Vijay Kumar: Flying Robots

Lex Fridman introduces Vijay Kumar as one of the top roboticists in the world, a professor at the University of Pennsylvania, dean of Penn Engineering and former director of the GRASP Lab, which was established in 1979, 40 years earlier. Kumar is best known for work on multi-robot systems, robot swarms and micro aerial vehicles, robots that cooperate in flight despite the uncertainty and challenges of real-world conditions.

Kumar's first robot, built when he was a graduate student, was part of a large project to build a hexapod weighing close to 7,000 pounds. It was powered by hydraulic actuation with 18 hydraulic motors, each controlled by an Intel 8085 processor and an 8086 co-processor, plus a 19th computer to coordinate the joints. His thesis work was coordinating the 18 legs and particularly the pressures in the hydraulic cylinders to get efficient locomotion.

Fridman asks how hard it was to make all the motors communicate. Kumar says the colleagues who worked on that wrote what he believes was the first multiprocessor operating system, in the 1980s, on computers with clock speeds of about half a megahertz.

Seeing the robot move was amazing, Kumar says, though in hindsight it should have been much smaller, as today's robots are, such as those of Boston Dynamics or Penn spinoff Ghost Robotics. At the time, he was constrained by available compute and substrate. As a matter of human psychology, bigger seems grander: people never appreciated nanotechnology the way they appreciate the Space Shuttle or the Boeing 747.

Asked for the most beautiful robot in motion he has seen, Kumar says what he is proudest of is his students' work on small UAVs that maneuver in constrained spaces and coordinate with each other to form three-dimensional patterns, which essentially creates 3D objects in the sky that can deform on the fly. He did a two-dimensional version earlier with ground robots, which was not as impressive. In 2011 the 3D version was a big accomplishment, although eight years later it is fairly standard.

Fridman asks whether beauty lies in distributed cooperation. Kumar says beauty to an engineer differs from beauty to an outsider. He finds beauty in shapes forming and deforming in midair, and in the agility of the individual components, and Fridman asks how quickly the shapes and components can be manipulated.

On terminology, Kumar dislikes the word drones because it signifies something dumb and preprogrammed that does one little thing, whereas robots are anything but. Unpiloted could mean radio controlled, so he thinks the right term is aerial robot. Agility is an attribute they do not have to have.

Asked which living creatures inspire his work, Kumar describes ants. Individually they are simple yet incredibly robust: an ant with six legs that loses one still works fine and may not even realize it. As a population they are incredibly resilient.