STEM

MIT Students Get a Taste of Programming for Autonomous Vehicles

• By Dian Schaffhauser
• 04/07/15

Every January for four weeks MIT students, faculty, staff and alumni have the chance to partake of "independent activities period" where they get to try out lectures series, how-to sessions, contests and other activities that strike their fancy. This January, a joint program offered by the Department of Aeronautics and Astronautics (AeroAstro) and Lincoln Laboratory gave participants an opportunity to learn how to program autonomous algorithms for driving cars without human intervention. The cars happened to be 1:10 scale, and the riskiest dangers they really faced were running into bystanders or the walls of MIT's subterranean tunnels, a network of passages that connects almost every building on campus.

Each team of four students was given an electric car with an on-board computer and a sensor suite. They attended seven lectures on algorithm robotics, including coverage of the robot operating system; algorithmic robotics, such as sensing, perception, control and planning algorithms; and case studies. Then they set to work preparing for the "Rapid Autonomous Complex-Environment Competing Ackermann-steering Robot (RACECAR)" race in a two-day hackathon. Their job was to build the software that would power the fastest robotic car in a friendly competition.

The robotic cars were modified to accept onboard control of steering and throttle actuators. Sensors perceived motion and the environment. The on-board computer running the student-coded algorithms and other software drivers — an NVIDIA Jetson Tegra K1 — processed the data generated by the sensors.

When race day arrived, three of four teams successfully completed the 515-foot course. The time of the fastest car was 49.64 seconds, about 7.1 mph.

According to the instructors, one reason to hold the course was to enhance embedded systems and robotics education at MIT. They're considering a repeat of the class in 2016, with the possibility of expanding the autonomous vehicle racecourse to a larger section of the institute's tunnel network.