This is a continuation of the Coffee Bot project. Now, instead of navigating via pads on the ground, the robot navigates between a series of reflective beacons. The idea of this is based heavily on the Laser Target Finder (http://www.philohome.com/sensors/lasersensor.htm )of Philippe Hurbain and I highly recommend his inspirational website (http://www.philohome.com/index.htm).
The Robot
The robot is shown below.
📷
The important feature for the navigation is the laser. This was taken from a standard laser-pointer and is mounted as closely as I could along the same optical axis as a standard light sensor (the connector of which is just visible coming out from under the RCX). When the laser strikes a catadioptric reflector target (see below) it is reflected back along the same optical path, and so directly into the light sensor. To distinguish between variations in background light intensity and reflected laser light the beam is pulsed by a rotating chopper and the software measures the peak-to-peak light intensity rather than the absolute value. In this way the robot was able to function in all light conditions.
📷 Front view of the robot, with the laser light just visible emerging under the RCX
📷
Photo taken from just above the robot showing the strong reflection from the target
The targets were simply beer cans (empty of course) with two strips of Scotch Reflexband tape
📷
A target
Locomotion for the robot is supplied by two motors each coupled with a rotation sensor.
In action the robot turns slowly until the target is found, then moves forward a certain distance. Once the robot had moved the appropriate distance it turns to face in approximately the right direction and then begins to search for the next target. To keep the robot on course between the targets it uses a kind of waddling movement with each motor moving the robot forward in turn, with the size of the movements regulated to keep the target it the centre of movement.
Performance wise this approach worked well, with the robot being able to reliably navigate through the apartment. However, it did suffer from several problems: it is slow both in finding the targets and in navigating between them; there is a tendency for navigation errors to accumulate.