Docking Logic - Chapter 3

On board Sensing - Infrared

Uploaded 7/18/07

 In this chapter we will discuss the optical sensing techniques that will be integrated into the robot itself, to enable it to locate and successfully dock with the charger beacon. There are two steps involved with this process. First, the robot must become aware that it is within range of the charger beacon with a non directional Omni type sensor that will detect the beacon no matter which direction it is relative to the robot. This is called "Proximity Detection". Once the robot has determined it is near the charger, a less sensitive narrow angle sensor allows it to home in on the beacon and direct it to the charging base. This is called the " Directional Sensor". Here we will cover both topics, and give you a sense of the approaches that are commonly used to get our bot to its base. In these diagrams, sensors are usually color coded red, as is the source IR beams.

Proximity Detection

 A very simplistic approach to proximity detection is to enclose a sensitive IR detector inside a translucent dome. When the surface of the dome is struck with infrared modulated light, the sensor will see the domes underside light up and a crude detection is accomplished. While fairly low in sensitivity, if one uses a Panasonic PN4602 integrated sensor, the range can be six feet or more in a normal room brightness.
 An overhead view of how this works is seen here. The beam will sweep over the robots central detector as the robot moves forward only when it is within the narrow confines of the beam. The robot would then immediately stop, and start its routine for determining the beams point of origin.
 With a centrally located multi-beam array on the beacon, the robots chances for detecting its proximity to the charger are greatly increased. Such an extreme approach is suitable for office or factory environments, however the single beam is more suitable for a home environment.
Directional Sensing
 This is how the basic analog and digital vane sensors work. When the vane points to the left of the beam, the right side is in the light and the left side is in the shadow of the vane and sees no light. By rotating the vane back and forth by rocking the body of the robot or moving the sensor on a servo motor, we can determine when the vane is directly aiming at the source when both sensors just see light at the same time. In the case of an analog system, cadmium sulfide photo cells can give a linear voltage reading which indicates the source direction by comparing the two output voltages. For a digital sensor such as the PN4602, when the vane is exactly pointed at the source, both will have an output. Proper filtering of the digital outputs is ESSENTIAL for giving the microcontroller a clean signal. More on this later.
 Another type of directional sensor used is the tube type. Here, a single photocell is mounted at the bottom of a black tube. Only when the tube is aimed directly at the source will there be an output. The robot then rotates 360 degrees slowly and when the beam is detected stops rotation. It then drives forward until it looses the beam, then does its search all over again. This simplistic approach can work fairly well, but the robot uses up a lot of energy spinning every few feet to get to the charger!
 A rotating cylinder with a single photocell in the center or back of the tube opposite of the slit can also be used to locate the beacon. The idea is either the entire robot or just the cylinder rotate, and when the slit in the side of the cylinder is aimed exactly at the source, the photocell will become active. I rather like this approach because the cylinder can be easily turned by a servo, and the slit can be so narrow as to get directional accuracy needed for a precise navigation to the charger.
 An example of a four way vane sensor in action. This was one of our experiments with the PicBot5 and provided first, an crude direction for the robot to turn, and second the vanes can be used with two of the sensors to home right in on the robot. Later we will have lots of images and some movies of this technique in action.
 With an analog approach, and two IR phototransistors, we can simply measure or compare the voltages from two sensors mounted about where the eyes are in the graphic on the right. You can either use a compartor to do the comparison and put out a digital signal to the processor or send both analog voltages directly to the analog in ports on the micro.
 A very cool approach to beacon location is to rotate the IR beam on the charger, and encode its rotation angle in the data stream on the IR beam. The robot then knows its polar angle from the charger and perhaps using sonar can get the radius to the charger to get its exact position in the room.
 The "Roomba" docking technique is to encode in the beam what the robot should do if it encounters the beam. So the robot will drive along, into one of the side beams and start turning toward the charger as it is instructed to do so. Such a system must be perhaps in a hall way to limit the robot so it does not cross the center beam first.
 Another thought at encoding the beams with data, if we simply number the beacon beams which are active all of the time, we can get directional information to the charger. This allows us to COMBINE a charger beacon with a complete navigation system in a small room or office area in one unit.


You are visitor number since June 17, 2001