Flying on Mars – or any other world – is an extraordinary challenge. An autonomous spacecraft, operating millions of miles away from pilots or engineers who could intervene on Earth, must be able to navigate an unknown and changing environment, avoid obstacles, land in uncertain terrain, and make decisions entirely independently. Each maneuver depends on careful perception, planning, and control systems that are fail-safe and allow the individual to recover if something goes wrong. A single miscalculation could cause a multimillion-dollar spacecraft to land face down on the surface, ending the mission before it even begins.
“This problem has not been solved in any way in industry or even in the research community,” says Nicholas Roy, the Jerome C. Hunsaker Professor in MIT’s Department of Aeronautics and Astronautics (AeroAstro). “You have to put together a lot of pieces of code, software, and integrate a lot of pieces of hardware. It’s not easy to put it together.”
This is not inconsequential, but for students approaching the culmination of their undergraduate career in Course 16, it is by no means impossible. In class 16.85 Autonomy Capstone (design and testing of autonomous vehicles), students design, implement, implement and test a full software architecture for flying autonomous systems. These systems have a wide range of applications, from urban air mobility and reusable launch vehicles to extraterrestrial exploration. With resilient autonomous technology, vehicles can operate far from home while engineers monitor the situation from mission control centers not too dissimilar to the high bay at AeroAstro’s Kresa Autonomous Systems Center.
Roy and Jonathan How, a Ford professor of engineering, developed the novel course based on class 16.405 (Robotics: Science and Systems), which introduces students to working with intricate robotic platforms and autonomous navigation in ground vehicles with embedded software. 16.85 applies the same principles to flight, using a basic quadrotor drone and a completely blank slate to build its own navigation systems. The vehicles are then tested on an obstacle course with questionable landing pads and uncertain terrain. Students work in vast teams (in the first case, two teams of seven – SLAMdunkers and Spelunkers) designed to mirror real-world missions where role coordination is necessary.
“Vehicles must be able to distinguish all hidden mission threats from the environment they are in and still survive,” How says. “We really want students to learn how to create a system that they can trust.”
Design and testing of autonomous vehicles
Video: MIT AeroAstro
Mission: Solve it together
“The specific mission we gave them this semester is to imagine that you are some kind of airplane that you have to fly and explore the surface of an extraterrestrial body such as Mars or the Moon,” Roy explains. “You must use onboard sensors to fly and explore, build a map, identify objects of interest, and then land safely on a surface that is probably not flat or perfectly horizontal.”
A mission of this size is too intricate for a single engineer to handle alone, but it also poses a challenge for a vast team. “The hardest problems right now are coordination problems,” says Andrew Fishberg, a graduate student in the Aerospace Control Laboratory and one of three teaching assistants (TAs) in the course. “To use the term robotics, a team of this size is something of a heterogeneous swarm. Not everyone has the same skill set, but everyone contributes something, and managing that together is a challenge.”
The challenge requires students to apply many types of “systems thinking” to the task. Relationships, interdependencies, and feedback loops are central to software architecture and are equally critical in the way students communicate and coordinate with their teammates. “Writing reports and communicating with your team sometimes seems like a lot of work, but if you don’t communicate, you’re a one-man band,” Fishberg says. “We no longer have ‘lone inventor’ situations where one person comes up with everything – it’s hundreds of people building this massive creation.”
Up-to-date faces of flight
Students taking the class say they are eager to step into a rapidly evolving field, working with unconventional tools and vehicles that go beyond time-honored applications.
“We still send rovers to extraterrestrial bodies. However, there is growing interest in deploying unmanned systems to study Earth,” says Roy. “There are many places on Earth that we want to send robots to explore, and it is dangerous for humans to travel there.” It is this growing set of applications that attracts students to work in the field.
“I was really excited about the idea of new classes, especially ones that focused on autonomy, because that’s where I see my career growing,” says senior Norah Miller. “This class gave me a really great experience in software development from scratch to a full flight mission.”
The Autonomous Vehicle Design and Testing course offers a unique perspective from instructors and technical experts who have known many of the students during their undergraduate careers. As a culmination, it gives you the opportunity to see how growth has come full circle. “We solved differential equations a few years ago, and now we’re implementing the software they wrote on a quadrotor at a high plant,” How says.
After weeks of learning, building, testing, refining, and finally flying, the results reflected the course goals. “It was exactly what we wanted to see,” says Roy. “We gave them a pretty difficult mission. We gave them equipment that should be able to complete the mission, but it’s not guaranteed. And the students put in a tremendous effort and really rose to the challenge.”