(SLAM) SIMULTANEOUS LOCALIZATION AND MAPPING I

Have you ever wondered why we don´t see robots everywhere? You may think technology isn´t mature enough, or they are too expensive to be built massively. You are wrong. We don´t see robots everywhere because they don´t know how to figure out where they are.

It may seems a simple task. We, as humans, only have to look around and get a general idea of where we are. But when we are talking about robots it´s not that simple. A robot must get information of the environment throug it´s sensors. Then this information (actualy only numbers or voltages), must be interpreted in some manner. A number of methods exists that make a computer or a robot able to extrac some information from images, or from laser range measurements, but most of them are computationally expensive, too much to be aplied in real time.

SLAM is a generic group of methods that try to solve the problem of localization (where am I in a given map?) and mapping (build a map of the environment where I am). Almos all state of the art SLAM methods are probabilistic. They try to model the uncentainties existing everywhere over the process. Robot movement (odometry), sensors readings, map belief, are some examples of the kind of uncertainties mentioned before. One of the most important researchers in mobile robotics and, particularly, in SLAM is Sebastian Thrun. Visit his webpage to get a huge amount of information (some technical background needed).

The day we completely solve the SLAM problem, we will see robots everywhere. An autonomous robot will be able to know where it is, and to recognise some elements in the environment necesaries to achieve it´s goals.