Learn More
Robots use a surprisingly simple but powerful algorithm to find out where they are on a map, a problem called localization by engineers. The algorithm known as particle filtering looks amazingly cool. In this first article, we attempt to explain the intuition behind particle filters. In part 2 we will elucidate the mathematics needed to build your own particle filters.
The Problem
Every robot that can move around, whether it is a vacuum cleaning robot like a Roomba or a self-driving car like Carla, has a lifelong problem to contend with. The problem is to find out its whereabouts on a map that every robot carries with itself. But why is finding the whereabouts so challenging? A robot can just ‘look’ at its surroundings and recall which area on the map looks like the current surroundings, not very different from how we humans find where we are in a city or inside the office when woken up from a post-lunch slumber.
The challenge is the utmost precision with which robots have to localize themselves, very unlike humans. If a Roomba thinks it is in front of a door, while it actually is slightly behind a wall, a few centimeters away from where it thinks it is, it may never be able to maneuver its way out from one room to another. A self-driving car, operating under a similar misconception, may scrape another car, veer off the road or climb a curb. The reason this does not happen is because robots are able to beat humans at one of their own games.
Humans try to address the imprecision in their beliefs by filling in missing clues using logic and reasoning but they do end up tripping sometimes. However humans move around rather slowly and have enough time to recover from a bad assumption. So, while we may miss a step on a staircase and stumble, in most cases, we do not tumble down the staircase into the abyss.
For the sake of everyone’s safety robots pack much more precision into their beliefs so that they do not have to trip and recover. How do robots synchronize their beliefs with reality so precisely? Let us find out, starting with something called priors.
Continue Reading
