This Soccer-Playing Robot Dog Can Dribble A Ball Across Sand, Mud, And Snow

Recently, China-based Hengzhi Future Team debuted Sparky, its iteration of a robot dog set to rival Boston Dynamic’s famed Spot. Now, a new pooch has joined the mix, as researchers from the Massachusetts Institute of Technology unveil ‘DribbleBot’.

As suggested in its name, this mechanical canine comes with a special ability to dribble a soccer ball under the same conditions as humans, if not better. According to scientists, the four-legged robot can hold its own across a series of terrains through onboard sensing and computing.

Be it a regular pitch of grass, or a trickier setup involving sand, gravel, mud, and snow, DribbleBot can adapt to the changing surroundings and the varied impact on how a soccer ball would travel. In fact, after every fall, the robot gets up and recovers the ball like a “committed athlete.”

Of course, it’s unlikely the robot dog will ever reach the expertise and skills of soccer greats Lionel Messi or Cristiano Ronaldo, but it still puts up an impressive display with its legs actually dribbling the ball along the ground.

Simulations were used to test the pooch’s ability across different terrains by creating a digital twin of the DribbleBot. Then, the robot’s assets and physical parameters were set, with four thousand versions of the simulation being run in parallel to collect as much data as possible.

Interestingly, just like an infant, the robot canine started off not knowing how to dribble a ball. The team trained it by giving it a reward when it succeeded, or negative reinforcement when it failed. It then got the hang of it in several days and continued to improve over time.

The system involved a set of sensors that allowed the robot to perceive its environment, “understand” its position, and “see” its surroundings, while its actuators enabled it to move itself and the ball forward, with the center “brain” tasked with converting sensor data into actions.

Naturally, controlling a soccer ball is a tougher feat than just walking, so the team made use of cameras attached to the DribbleBot’s head and body to give it a new sensory modality of vision in addition to its motor skills.

Researchers believe developing non-wheeled robots is a step toward the future. For example, in the case of a natural disaster, such as a flood or earthquake, these machines would be better suited to aid humans in search-and-rescue operations over varying terrains.

“The whole point of studying legged robots is to go terrains outside the reach of current robotic systems. 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,” said Professor Pulkit Agrawal.

[via Hackster.io and TechCrunch, cover image via Mike Grimmett / MIT CSAIL]