VIDEO: MIT Cheetah Robot Reaches Major Milestone

Device becomes the first four-legged robot to jump over obstacles autonomously

MIT engineers have accomplished a giant leap for robot development – figuratively and literally. The researchers behind the robotic cheetah have given it the ability to “see” and jump over obstacles as it runs. According to the team, this is the first time in the history of robotics that a four-legged device has been able to do so autonomously.

So how does it jump over hurdles? The robotic cheetah behaves similar to a human runner and plans out its path. Once it detects an approaching obstacle, the robot is capable of estimating the object’s distance and height.

Testing the robot 

The team tested its robot both on an indoor track and on a treadmill. The device successfully jumped over obstacles measuring up to 18 inches (approximately half of the robot’s height). It managed a running speed of 5 mph throughout the process.

“A running jump is a truly dynamic behavior,” says Sangbae Kim, an MIT assistant professor of mechanical engineering. “You have to manage balance and energy and be able to handle impact after landing. Our robot is specifically designed for those highly dynamic behaviors.”

Earlier this year, the engineers allowed the robot to run untethered. At that point, it didn’t have cameras or any other vision system. Thanks to LIDAR (a vision system that maps terrain with the help of reflections from a laser), the robot can now “see.” The LIDAR data helped the team create an algorithm to plan out the robot’s path, which gives the device autonomous control.

Integrating an algorithm 

The algorithm has three components, the first of which allows the robot to detect an obstacle and approximate its size and how far it is. A formula essentially simplifies visual scenes for the robot; the ground is represented by a straight line and obstacles are represented as deviations from that line, which makes it easier for the robot to calculate the height and distance of hurdles.

The second component of the algorithm comes into play once the robot detects an obstacle. It uses the obstacle’s distance to predict the most suitable position to jump and clear the hurdle. It either speeds up or slows down the robot to achieve the best jumping-off position. This optimization component, dubbed the “approach adjustment algorithm,” takes approximately 100 milliseconds to complete.

Feasible vs. optimal 

Finally, the algorithm’s third component, which serves to determine the robot’s jumping trajectory, kicks in when it starts to leap. The researchers created a formula that calculates the amount of force required from the device’s electric motors in order to launch the robot in a safe manner.

 According to Kim, the algorithm provides a feasible jump as opposed to an optimal one. “If you want to optimize for, say, energy efficiency, you would want the robot to barely clear the obstacle — but that’s dangerous, and finding a truly optimal solution would take a lot of computing time,” Kim says. “In running, we don’t want to spend a lot of time to find a better solution. We just want one that’s feasible.”

The team will present its developments at the DARPA Robotics Challenge in June, along with a paper about the autonomous system the following month at the Robotics: Science and Systems conference. For more information, visit MIT’s website.

Thumbnail photo credit: Haewon Park, Patrick Wensing, and Sangbae Kim.