Manipulate the future | MIT News


As robots evolve, society’s collective imagination always wonders what else robots can do, with recent fascinations coming to life as self-driving cars or robots that can walk and interact with objects like humans do.

These sophisticated systems are fueled by advances in deep learning that have sparked breakthroughs in robotic perception, so today’s robots have greater potential for better decision-making and improved functioning in real environments. But tomorrow’s roboticists need to understand how to combine deep learning with dynamics, controls, and long-term planning. To maintain this momentum in robotic manipulation, today’s engineers must learn to soar above the whole field, connecting an increasingly diverse set of ideas with the interdisciplinary approach needed to design increasingly complex robotic systems.

Last fall, MIT’s Department of Electrical Engineering and Computer Science launched a new course, 6,800 (Robotic Manipulation) to help engineering students study the latest advances in robotics while solving real industry problems. This is a unique course that can provide a foray into robotics for students without any robotics background, designed by Russ Tedrake, Toyota Professor of Electrical Engineering and Computer Science, Aeronautics and Astronautics, and Engineering mechanics at MIT. Tedrake developed the course after robotic manipulation became the new focus of his own research at the Toyota Research Institute and the Robot Locomotion Group at MIT, and it became apparent to him that developing an educational framework would be important because the field is so diverse and changes so quickly. .

“It was about time there was one place where you could really see all the rooms,” says Tedrake.

Its students learn fundamental algorithmic approaches to building robotic systems capable of autonomously manipulating objects in unstructured environments. Exploring topics such as perception, planning, dynamics, and control, students solve problem sets to guide themselves in developing a software stack, typically using the permissive license. Drake open-source software – that’s why industry leaders also watch Tedrake’s conferences. Not focused on quizzes and final exams, the course instead culminates in a final project where students can explore any problem in robotic manipulation that fascinates them.

David von Wrangel, an undergraduate engineering student, is currently taking the course. He has a background in rocketry and propulsion, and he’s only interested in robotics thanks to a recent mobile robotics internship at Tesla. Upon learning what it takes to make a robot move, his next question soon became: how do you get a robot to pick up something? That’s when a colleague from his internship told him about MIT’s Robotic Manipulation course.

“I was super excited, because that was exactly what I was missing: now that you can manipulate your robot, you just have to figure out how to use manipulation to move other things,” says von Wrangel.

At Tedrake course notes providing students like von Wrangel with a window to observe the professor’s mental landscape in the field. Students say they’re unlike any lecture notes they’ve ever seen — providing a constantly updated roadmap of what it would take to advance robotics as a field.

Teaching assistants (TAs), like doctoral student HJ Terry Suh last fall, develop problem sets that provide opportunities for students to apply less familiar concepts and see for themselves how various disciplines connect. .

This potential to gain a new point of view in robotics is precisely what attracted graduate student Anubhav Guha to the course. His research focuses on applications of controls, and he took the course to examine open problems in robotics directly related to his research. “I kind of wanted to explore the field a bit and get a sense of the technical issues,” Guha says.

Tedrake says the industry is increasingly interested in skilled handling engineers, and this demand motivated him to launch the course. “Manipulation is exploding in the field,” Tedrake says, adding that recently, “it’s less of a niche area, everyone is looking to get robots to do things with their hands.” Right now, big companies are investing.

They are not the only ones investing in the future of robots. Because there was so much student interest in the course, Tedrake decided to open registration for robotic manipulation to undergraduate and graduate students.

For doctoral student Daniel Yang, who took the course when it was first offered last fall, his interest in robotic manipulation grew from working in industry and seeing the limited environments in which robots are currently working. As part of the joint MIT/Woods Hole Oceanographic Institution (WHOI) program, Yang collaborates with oceanographers, helping to build autonomous robotic systems that can dive into water to collect scientific data.

“Overall, I’m interested in bringing robots out into the real world,” Yang says.

For Yang’s final course project, he was inspired to explore the inner workings of a tossing robot like Google’s TossingBot. By collaborating with a partner to build their own ball-throwing robot in a simulator, they were able to clearly document the effects that threw the robot out of its throwing game, as the course simulator allowed them to freeze and re-examine each interaction during troubleshooting. For example, they realized that the accuracy of their robot’s throwing arm was limited because the exact physics of how the ball interacted with the robot’s gripper was still unknown.

“You might think that picking up something with your fingers is quite simple,” Yang says. “But when you try to translate that into simulation, there’s so much added complexity.”

Suh says Guha’s final project was among the most ambitious. Guha has created a complete simulated manipulation system that puts together a puzzle using a camera system that detects the correct image placement and orientation for each piece of the puzzle.

“Some other projects kind of focused on one aspect of the manipulation pipeline, like input or perception,” Guha says. “And they really went deep into it. And I wanted to explore all the different components needed to create a fully functional system.

According to Tedrake, whether students decide to focus on one aspect of the manipulation pipeline or tackle an entire system, problem solving at all levels equates to great success in this rapidly growing and thirsty field. solutions.

“Even if it’s an algorithm that I’m familiar with, but seeing what they struggled with or how they succeeded, it refines my understanding of the algorithm,” Tedrake says.

Tedrake takes many of the solutions found in his classroom right into the lab, advancing his own research and gaining new research ideas every week. And sometimes he takes the students with him. That’s what happened with von Wrangel, whose enthusiasm for the course led Tedrake to recruit him to help refine the algorithms for Tedrake’s Robot Locomotion group.

Yang and Suh agree with Tedrake that the future of robotic manipulation is just around the corner and that the robotic manipulation course will help train engineers to stabilize the field as it progresses.

“I think in the last few years there have been so many advances in all these different areas, but there’s nothing that ties them together in a specific problem space,” Yang says.

Suh thinks people will start to see more advancements in everyday life, and as that happens, the robotic manipulation course will be there as a resource. His vision of the future sees robots everywhere.

“It will be a kind of manipulation in nature, where we will have robots that will enter arbitrary places, such as people’s homes or kitchens, and perform very delicate manipulation tasks that we generally expect from humans,” says Suh. .

For von Wrangel, the future of robotic manipulation will help us move far beyond the apparent luxury of self-driving cars and robot bunnies. “I’m very interested in space exploration and humans becoming multiclimate, multiplanetary species,” von Wrangel says. “And I believe robots could help us build our future on Mars.”

In its second year, Robotic Manipulation continues to inspire bigger, bolder ideas from students, who ask wizards like Suh sometimes esoteric questions in the middle of the night, always looking for guidance on how to better simulate systems. . Tedrake says the ideas for this year’s final projects have already been more ambitious than last year. Each project, tweaking each algorithm little by little or imagining entire systems from a unique point of view, will play a driving role in innovation in the field.

“Some years, with student projects, it’s more like, I want to try this paper and just make it a little bit better,” Tedrake says. “And this year, it’s like, I want a robot that will tie my shoes.”


Comments are closed.