CHANNELS

Aerospace

Automotive

Automated & Autonomous Vehicles

Battery & Electrification Technology

Defense

Drone TV

Electronics

Manufacturing & Prototyping

Materials

Medical

Motion Control

RF & Microwave Technology

Robotics & Automation

Sensors & IoT

Test & Measurement

Aerospace
Automotive
Automated & Autonomous Vehicles
Battery & Electrification Technology
Defense
Drone TV
Electronics
Manufacturing & Prototyping
Materials
Medical
Motion Control
RF & Microwave Technology
Robotics & Automation
Sensors & IoT
Test & Measurement
Aerospace & Defense
Search Results

Meet 'DribbleBot,' The Next Great Soccer Star

MIT researchers have developed a four-legged robot capable of playing soccer like a human. See how this robotic system maneuvers a soccer ball on various terrains and may even aid humans in search-and-rescue missions.

"Our goal in developing algorithms for legged robots is to provide autonomy in challenging and complex terrains that are currently beyond the reach of robotic systems," says Pulkit Agrawal  , MIT professor, CSAIL principal investigator, and director of Improbable AI Lab.

Topics:
Automation Robotics

More From SAE Media Group

    Transcript

    00:00:00 (upbeat rock music) - The mission of Improbable AI Lab is to build robots which can perform the same physical tasks that humans perform today. The things that we want robots to be doing so humans can focus on what they are good at, and what they want to do to realize their full potential. In simulation, our robots are learning by trial and error. They're essentially trying to figure out what sequence of commands should they give to their legs.

    00:00:40 One fun task is for these robots to play soccer, and definitely there has been a lot of work in robots playing soccer, but a lot of this work has been with robots which are not as agile or dynamic. - Legs robots are really interesting to work with because they're a bit challenging to control. To dribble a soccer ball, it's not sufficient for the robot to rely on what it feels through its motors.

    00:01:04 It also needs to know where the soccer ball is, and there's not enough information about that in just what the robot's feeling. It needs vision. It needs to be able to perceive the ball from a distance and run to it and also kick it when it's sort of underneath the robot, even if it's moving in a way that's not expected. For this, we actually do leverage the cameras in the robot's head and body to detect and perceive the ball

    00:01:28 and make decisions based on that information. Our robot's having a problem, which is when it kicks the ball on grass, the ball slows down and it recaptures the ball and it keeps dribbling. But when it kicks the ball on pavement, the ball just keeps you rolling forever, because it doesn't slow down. So we implemented a drag model to model how the robot slows down on grass,

    00:01:50 but we need to modify it so that the robot is capable of catching up when the ball's rolling fast on pavement. I think it's fun to watch a human interact with the robot, and the soccer ball is a really good medium for that. - So what you're going to see in many of these videos is robots dancing, going under things, going over obstacles, running on ice, many different things. We don't have to go to each task and have a human do our design strategies which are specific to that task.

    00:02:20 All of these strategies are a result of the same way of training the robot. We train in simulation by trial and error, and then deploy them in the real world. - We're excited to continue our research on legged robotics to unlock their potential to accomplish dynamic tasks in their environment and assist humans.