In a leap for robot development, the MIT researchers who built a robotic cheetah have now trained it to see and jump over hurdles as it runs — making this the first four-legged robot to run and jump over obstacles autonomously.
To get a running jump, the robot plans out its path, much like a human runner: As it detects an approaching obstacle, it estimates that object’s height and distance. The robot gauges the best position from which to jump, and adjusts its stride to land just short of the obstacle, before exerting enough force to push up and over. Based on the obstacle’s height, the robot then applies a certain amount of force to land safely, before resuming its initial pace.
In experiments on a treadmill and an indoor track, the cheetah robot successfully cleared obstacles up to 18 inches tall — more than half of the robot’s own height — while maintaining an average running speed of 5 miles per hour.
“A running jump is a truly dynamic behavior,” says Sangbae Kim, an assistant professor of mechanical engineering at MIT. “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.”
Kim and his colleagues — including research scientist Hae won Park and postdoc Patrick Wensing — will demonstrate their cheetah’s running jump at the DARPA Robotics Challenge in June, and will present a paper detailing the autonomous system in July at the conference Robotics: Science and Systems.
Watch the MIT researchers' robotic cheetah take a running leap over obstacles.
Video: Haewon Park, Patrick Wensing, and Sangbae Kim
See, run, jump
Last September, the group demonstrated that the robotic cheetah was able to run untethered — a feat that Kim notes the robot performed “blind,” without the use of cameras or other vision systems.
Now, the robot can “see,” with the use of onboard LIDAR — a visual system that uses reflections from a laser to map terrain. The team developed a three-part algorithm to plan out the robot’s path, based on LIDAR data. Both the vision and path-planning system are onboard the robot, giving it complete autonomous control.
The algorithm’s first component enables the robot to detect an obstacle and estimate its size and distance. The researchers devised a formula to simplify a visual scene, representing the ground as a straight line, and any obstacles as deviations from that line. With this formula, the robot can estimate an obstacle’s height and distance from itself.
Once the robot has detected an obstacle, the second component of the algorithm kicks in, allowing the robot to adjust its approach while nearing the obstacle. Based on the obstacle’s distance, the algorithm predicts the best position from which to jump in order to safely clear it, then backtracks from there to space out the robot’s remaining strides, speeding up or slowing down in order to reach the optimal jumping-off point.
This “approach adjustment algorithm” runs on the fly, optimizing the robot’s stride with every step. The optimization process takes about 100 milliseconds to complete — about half the time of a single stride.
When the robot reaches the jumping-off point, the third component of the algorithm takes over to determine its jumping trajectory. Based on an obstacle’s height, and the robot’s speed, the researchers came up with a formula to determine the amount of force the robot’s electric motors should exert to safely launch the robot over the obstacle. The formula essentially cranks up the force applied in the robot’s normal bounding gait, which Kim notes is essentially “sequential executions of small jumps.”
Optimal is best, feasible is better
Interestingly, Kim says the algorithm does not provide an optimal jumping control, but rather, only a feasible 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.”
Sometimes, that means the robot may jump much higher than it needs to — and that’s OK, according to Kim: “We’re too obsessed with optimal solutions. This is one example where you just have to be good enough, because you’re running, and have to make a decision very quickly.”
The team tested the MIT cheetah’s jumping ability first on a treadmill, then on a track. On the treadmill, the robot ran tethered in place, as researchers placed obstacles of varying heights on the belt. As the treadmill itself was only about 4 meters long, the robot, running in the middle, only had 1 meter in which to detect the obstacle and plan out its jump. After multiple runs, the robot successfully cleared about 70 percent of the hurdles.
In comparison, tests on an indoor track proved much easier, as the robot had more space and time in which to see, approach, and clear obstacles. In these runs, the robot successfully cleared about 90 percent of obstacles.
Kim is now working on getting the MIT cheetah to jump over hurdles while running on softer terrain, like a grassy field.
This research was funded in part by the Defense Advanced Research Projects Agency .
Comments
Hilary Albutt
May 29, 2015
now what would be quirky but cool would be to clad it in some star wars costume. two come to mind..
Kiruby
May 29, 2015
Awesome!
Frederick Murre
May 29, 2015
Impressive progress.
The big question I have is, why do all of these designs have a rigid spine?
I can understand if you're trying to isolate leg behaviors better so you can study them- ie, start with smaller systems.
A flexible spine helps with extension, sustained obstacle clearance, and shock aborption, and direction changing in actual running mammals.
AT
May 29, 2015
Just be sure to leave it with a couple weaknesses for when it's eventually sicked on humanity.
llama256
May 29, 2015
Well the Laser scanner makes it a heck of a lot easier than a vision system.
Peter Mokhothu
May 30, 2015
Groundbreaking. I can only imagine how useful this robot could become. All the possibilities. However, wouldn't it be better for it to make a leap before reaching an obstacle, in order to detect the length of it and then make the leap after that? I'm sure the plan is not to make it jump over thin objects only.
zhaozj89
May 30, 2015
Why not add a tail to keep its balance, like real animals do?
Push Bhatkoti
May 30, 2015
this is just the start.. i wonder if there is a source code for this project and open for developer. i can't find the codes on MIT site.
LoQuan Seh
June 1, 2015
Can we rent or buy these robots for events?
Ταξικος
June 2, 2015
what is useful in this for humanity?
DORSIGUER
June 2, 2015
Is machine learning used on this?
Seems to me that it would save a lot of the developer's time specially if you want to introduce future obstacles.
Laurent Pequin
June 3, 2015
So Philipp K. Dick will be right soon : "Do Androids Dream of Electric Sheep?"
Ruben Pablos
June 4, 2015
It is the second logical phase of this project
https://youtu.be/XMKQbqnXXhQ
Tin Cup NYC
June 5, 2015
This is great. Congratulations. Cheetahs run very fast, in sprints, and are not long distance runners. With this robotic cheetah, the MIT scientists have solved a genetic problem!
Caden K.
June 14, 2015
cool!awesome!
trew poiu
July 5, 2015
cool
Darien Craig
September 5, 2015
I love their filming solution, with the cameraman being pushed in a box.