Robot Vacuums Can Finally Climb Stairs, It Only Took 60 Years Since Shakey the Robot

In 1966, a team at the Stanford Research Institute in Menlo Park, California, built a robot that could see, think, and move. They named it Shakey, because it shook when it moved. Funded by DARPA, Shakey was the first mobile robot that could reason about its actions, plan routes through rooms, and push objects around obstacles.

Shakey was the size of a refrigerator, mounted on wheels, equipped with a TV camera and bump sensors. It communicated with a mainframe computer via radio link. Processing a single movement command could take hours. The robot moved slowly, deliberately, and with great uncertainty through a specially constructed environment of simple rooms and colored blocks.

The project ran from 1966 to 1972. It produced fundamental advances in AI, including the A* search algorithm and the Hough transform for computer vision, that are still used today. But Shakey itself was impractical. A research curiosity. A million-dollar machine that could push a block across a room.

Sixty years later, at CES 2026, multiple companies demonstrated robot vacuums that can climb stairs.

“It took 60 years to go from 'a robot that can push a block across a room' to 'a robot that can vacuum your stairs.' The gap between research demo and consumer product is measured in decades.”

This sounds like a modest achievement until you consider what stair-climbing requires. The robot must detect the stair edge, assess the height and depth of each step, adjust its wheel geometry or leg mechanism, maintain balance during the transition, continue mapping its environment on a different floor, and resume cleaning, all autonomously.

The journey from Shakey to stair-climbing vacuum took six decades because each step required a different breakthrough. Shakey proved robots could navigate. The SLAM algorithm, formalized in 1986, let robots map unknown environments while moving through them. iRobot's Roomba in 2002 proved consumers would buy robots for housework, 30 million units sold by 2020.

Each generation added capability. Random bounce navigation gave way to systematic path planning. Infrared sensors gave way to cameras. Cameras gave way to LiDAR. By 2024, robot vacuums could identify and avoid pet waste, pick up socks with mechanical arms, and empty their own dustbins.

“Shakey's team at SRI spent six years teaching a refrigerator-sized robot to navigate a single room. Your $300 Roomba does it in minutes.”

But stairs remained the frontier. The multi-floor home was the robot vacuum's kryptonite. Most households solved this by buying a second vacuum for the upstairs, a hack, not a solution.

The CES 2026 stair-climbing vacuums use articulated wheel systems or leg-wheel hybrids that shift geometry for steps. Some use tracks like miniature tanks. All use AI-powered depth perception to detect stair edges in real time.

The technology isn't perfect yet. The demos were careful, controlled environments, specific stair types, modest heights. Mass production is likely a year or two away. But the principle has been proven.

The gap between Shakey and a stair-climbing Roomba is the gap between research and product. Shakey's team at SRI were exploring fundamental questions about machine intelligence. They didn't care about vacuuming. But the navigation algorithms they invented, A* pathfinding, SLAM, hierarchical planning, are running in the $300 vacuum that cleans your kitchen floor.

The root of your robot vacuum isn't iRobot. It's a shaky refrigerator-sized robot in a California lab, pushing colored blocks across a room in 1966, funded by the same agency that funded the internet.