How did Boston Dynamics teach Atlas to perform the Ghost Rabona kick?

The company used motion capture from a skilled soccer player, fed that data into a simulation engine, and applied reinforcement learning to optimize the motion without human-form constraints. The training took only a few hours to a day, much faster than older techniques would have allowed.

What are Spot robots doing at World Cup stadiums?

Spot is walking around stadiums looking for hazards and suspicious packages, equipped with a 360-degree camera, 4K high-definition imaging, thermal imaging, and lidar. Security teams operate the robots themselves, either autonomously or via teleoperation, to do rounds when most people are gone.

Why did Boston Dynamics switch to reinforcement learning for robot control?

Reinforcement learning made it possible to solve problems—like moving a fridge or performing complex footwork—that would have been difficult or impossible using the company's earlier model-based predictive control techniques. It also lets robots learn in simulation and apply that learning directly to physical robots.

RoboticsThe Robot ReportJul 6, 2026Jul 7, 2026, 06:00 JST

Boston Dynamics' Atlas learns complex soccer tricks using AI, Spot patrols World Cup stadiums

Send on LINE

3 Key Points

What happened
Boston Dynamics has taught its Atlas humanoid robot to perform the 'Ghost Rabona' soccer kick as part of Hyundai's partnership with FIFA World Cup. Spot quadrupeds are deployed at two stadiums in Dallas and two at Citi Field to support security operations by patrolling for hazards and suspicious packages.
Why it matters
The company shifted from classical predictive control to reinforcement learning (a machine-learning technique that lets robots learn behaviors in simulation), allowing Atlas to master complex interactions with objects and terrain that would be harder to program manually. The same techniques enabling Atlas to carry a fridge are now being applied to industrial warehouse tasks. For stadium operators, Spot provides autonomous or remote-controlled inspection without facial-recognition capability, complementing human security teams.
What to watch
The company emphasized that Atlas is far from matching human soccer players—it cannot yet interact safely with other people or run beside them. Spot is designed as a working tool for data collection and hazard detection, not a public spectacle; security teams themselves operate the robots in both autonomous and teleoperated modes.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, in your inbox every morning.

Get news like this every morning, free

Context & Analysis

Boston Dynamics' World Cup partnership marks a shift in how the company demonstrates its robots' capabilities and where they operate. Rather than competing in robot sports leagues, Hyundai is using the global platform to show Atlas and Spot performing real work—athletic tricks that require interaction with objects and terrain, and security patrols in crowded public venues. The technical foundation for this pivot is reinforcement learning, a machine-learning approach that trains robots in simulation by rewarding desired behaviors rather than programming each motion by hand. This method proved faster and more robust: the Ghost Rabona kick took only hours to a day to train, compared to the months of manual engineering that earlier techniques would have required. Critically, the same algorithms now enable Atlas to handle tasks like carrying a fridge, suggesting the company sees these showcase moments as proof-of-concept for industrial applications.

The deployment of Spot at stadium security represents a practical expansion beyond industrial settings (mines, nuclear facilities, warehouses). By having security teams operate the robots themselves rather than relying on specialists, Boston Dynamics is testing whether legged robots can integrate into existing workplace routines. The company was careful to note that Spot does not have facial-recognition capability, addressing privacy concerns head-on. However, Boston Dynamics also stressed that Atlas remains far from human soccer performance—it cannot yet safely interact with or move alongside people. This transparency suggests the company is managing expectations while building public familiarity with robots in everyday environments, a precondition for broader commercial adoption.

FAQ

How did Boston Dynamics teach Atlas to perform the Ghost Rabona kick?: The company used motion capture from a skilled soccer player, fed that data into a simulation engine, and applied reinforcement learning to optimize the motion without human-form constraints. The training took only a few hours to a day, much faster than older techniques would have allowed.
What are Spot robots doing at World Cup stadiums?: Spot is walking around stadiums looking for hazards and suspicious packages, equipped with a 360-degree camera, 4K high-definition imaging, thermal imaging, and lidar. Security teams operate the robots themselves, either autonomously or via teleoperation, to do rounds when most people are gone.
Why did Boston Dynamics switch to reinforcement learning for robot control?: Reinforcement learning made it possible to solve problems—like moving a fridge or performing complex footwork—that would have been difficult or impossible using the company's earlier model-based predictive control techniques. It also lets robots learn in simulation and apply that learning directly to physical robots.

The Robot ReportRead Original Article

Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Get it free

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.