The small four-legged robot LittleDog, from Boston Dynamics, has acquired an impressive array of improved locomotion skills thanks to researchers at the University of Southern California. The scampering robot shows off some of these skills in a new video, performing deft maneuvers to overcome obstacles and using machine-learning to plan its steps over tricky surfaces.
While its larger counterpart BigDog can recover from unexpected obstacles, like sliding on ice, LittleDog has to more cautiously plan its step to carefully but quickly move over rough, unfamiliar terrain.
DARPA introduced the 5-inch-tall robot a few years ago when it began its robot locomotion initiative, asking several universities to improve LittleDog’s learning, control, environment perception and locomotion. At about 5 pounds, LittleDog uses a host of sensors and three motors in each of its four legs, a camera and a machine-learning algorithm to find good footholds autonomously.
The video below from USC shows LittleDog walking autonomously in real-time, successfully navigating rocky terrain, a staircase, and performing special moves to get over barriers and avoid a gap. The program also lets it recover when it falls or stumbles. Neat stuff.
Smaller design teams can now prototype and deploy faster.