Robot Mapping

We think we're pretty clever the way we create drawings these days. Laser measurement devices and computers have made the creation of drawings an efficient process. The resulting CAD files are much more reliable than anything that was created using manual measurement and drafting techniques. The electronic files are very versatile and can be reused for a variety of applications.

So what's next?

I've found some great videos of robots measuring buildings. Look at this one for example:

And here's a guy adding some intelligence to his Roomba:

I have a Roomba at home and its not that smart. It has a problem with doors. It goes into the bathroom. It vacuums behind the door which causes the door to shut and then it's stuck in there. It bounces around, beeps and clicks for an hour and then runs out of power. The exhausted machine then has to be carried back to the docking station.

But, given the rapid development of technology, it is inevitable that robots will be able create CAD surveys of buildings within the next few years. But first, there are two technical problems that must be solved:

1. Robots must be able to safely navigate through buildings, not get lost or stuck in corners. This is not a big problem. It will occur through an incremental refinement of existing technology.
   
2. Robots must understand what they see. The survey robots that exist today can scan a space and create a point cloud of their environment. To create useful drawings the robot must be able to distinguish a door from a niche and a wall from a window. Electronic drawings and models have to indicate what things are, not just their outlines. Recognizing objects is a bigger problem that will require more computing power and clever programing.
   

By the time we have these survey robots, we will be used to seeing cleaning robots, maintenance robots and delivery robots. All these applications rely on the same two problems of spatial navigation and object recognition.