BEFORE — The chicken-and-egg problem

@thedataandaiguy

4.5K followers · usually 2.0K views a post

Views

8.3K

Likes

229

Comments

4

Shares

4

SoundTikTok sound · original sound - thedataandaiguy

Promoted product

Roomba

The Roomba® robot vacuum can give you the perfect cleaning solution for your home. See why Roomba® robot vacuums are the best for your floors.

2 videos2 creators

Why it worked

The video effectively explains a complex technical concept (Visual SLAM) using a relatable everyday object (Roomba), making advanced technology accessible and interesting to a broad audience.

Summary

The video explains how Roomba vacuums use Visual SLAM (Simultaneous Localization and Mapping) to navigate and map homes, overcoming the chicken-and-egg problem of needing a map to know location and vice-versa. This technology, first implemented in 2015, uses cameras, IMUs, and wheel odometry, processed by a Kalman filter, to create a coherent map.

Structure

1Introduction to the chicken-and-egg problem in mapping
2The old workaround with random bumping
3Introduction of Visual SLAM in Roomba 980
4

Original caption

Your Roomba runs the same algorithm class as a self-driving car. It's called SLAM — Simultaneous Localization and Mapping — and iRobot started shipping it in 2015 with the Roomba 980. #techtok #hereswhatyouneedtoknow #roomba #slam #LearnOnTikTok

More from @thedataandaiguy

Product placement

Roomba (acts: cleans floors, maps houses, uses SLAM algorithm; removing it would change the video's subject) iRobot (appears: brand name on Roomba; removing it would not change the video's content)

On-screen text

BEFORE — The chicken-and-egg problem To build a MAP -> need to know WHERE you are circular dependency unsolvable one-at-a-time To know WHERE you are -> need a MAP the old workaround (2002) your living room robot Roomba (2002) HOW — Visual SLAM on a Roomba (j-series / s-series, since 2015 Roomba 980) Ceiling-facing CAMERA -> visual features (corners, light fixtures) IMU -> acceleration + rotation Wheel odometry -> distance traveled KALMAN FILTER (some systems use a Particle Filter) - fuses noisy signals each pass refines both you are here one coherent estimate

Transcript

Your Roomba runs the same algorithm class as a self-driving car. It's called SLAM — Simultaneous Localization and Mapping — and iRobot started shipping it in 2015 with the Roomba 980. The first Roombas from 2002 bumped around at random. And in 2015, iRobot shipped the Roomba 980 with a camera on top. And the newer J-series and S-series now build an actual map of your house using visual SLAM, which stands for simultaneous localization and mapping. And the reason this is hard is it's a chicken-and-egg problem. To build a map, the robot needs to know where you are. But to know where you are, it needs a map. SLAM solves both at once. The camera picks out visual features like ceiling corners and light fixtures. The IMU tracks acceleration and rotation. Wheel odometry tracks distance traveled. And all that feeds into a Kalman filter, or in some cases, a particle filter, which then fuses the noisy signals into one coherent estimate that improves every pass. So if you're a practitioner, this pattern is everywhere. It's in drones, warehouse robots, even vision pro. All the same math, which is 40 years old, but it's also fitting in a $30 chip under the coffee.