Servo - December 2016

I know my wife would love it if a robot: • Could vacuum (we have a Roomba, but I need to fix it) • Could pick up small dropped items (that would choke a Roomba) • Automatically do the above when there is no one around I’d love it if a robot could fetch me snacks and drinks during hockey games, but that will take a much larger, smarter, and more powerful bot than Berry.

In order to fetch and move items, not only would the robot need to be much more powerful and have at least one arm, it would: • Have to know where it was in the house • Have to know the location of items of interest (in 3D space due to tables and shelves)

We have to learn to walk before we can run, so let’s forget a large robot with an arm for now, and start to work on having a robot be able to tell where it is in the house. This requires indoor positioning. There are many techniques for indoor positioning, including: • Measuring wireless signal strength (from Wi-Fi, RF beacons, RFID, etc.) • Looking for known optical beacons (whose positioning is known) • “Absolute” positioning from wiring or other position markers in the floor • Visual recognition of a camera view • “Dead reckoning” by starting from a known position, combined with odometry or gyroscope If you are interested in the MANY methods people have used, Googling “robot indoor positioning” will get you more links than you can shake a stick at.

For my robots, I don’t need extremely precise positioning. I will be happy with knowing where Berry is, say, within four inches. Currently, I am using: • A compass to a bearing of “magnetic north” • Correcting for magnetic declination so I can get “true north” • An accelerometer to compensate for compass tilt The above gives me a reasonably accurate indication of where true north is relative to where Berry is facing.

Combining knowing where north is with the distance to the closest object from it, Berry will be able to build a rough map of the floor he is running around on. Such a map is the first step to later being able to figure out WHERE he is — based on direction, distance, and a predefined map of the floor that he can try to locate himself on via pattern matching.