IR Triangulation

In this approach, an IrED is also used to shine a focused beam of modulated IR. Unlike the previous approach, however, the sensor is not a single phototransistor. Typically, the sensor is a linear array of CMOS or charged-couple devices. Due to the triangulation angular difference of reflection at different distances, different individual sensor cells sees the illuminated spot. This is then converted either to a digial or an analog signal as output.

This approach is not prone to varying reflectance of surfaces because it is not intensity-based. Furthermore, this type of sensor is self-contained with the lens and sensors. You do need to provide either a digital or an analog interface to read the distance, however. Furthermore, most of these sensors are consistent but not calibrated. This means even though a particular sensors provides the same reading at the same distance all the time, it may not be the same reading as another sensor.

Interpreting the results of IR triangulation is not straightforward. The distance is not linear to the output of the sensor. A translation table is usually needed if you need to know the distance. Even when a distance is available, you still need to know which wall reflects the IR based on the distance. Be prepared to spend some time on sensor interpretation logic if you are using IR triangulation sensors.

You can use this type of sensor for both error correction and mapping. For error correction, you should consider arranging the sensors so that they are not at 0, 90 and 270 degrees from the direction of motion. Instead, you need two sensors that sense at an angle so they can sense the side of walls that are ahead of the robot. Refer to figure 13.1 for illustration.

Figure 13.1: Configuration of IR rangefinding sensors

$$\includegraphics{robot-ir-rf}$$

The typical range of this type of sensor is from 100mm to 10000mm. This is a very useful range for micromouse applications. In addition, the precision (based on triangulation) is much better at shorter distances, which means such sensors are useful for both error correction and mapping.

This type of sensor is not cheap. An analog sensor typically costs $8 to $15. However, one sensor can replace many individual sensors in other types of designs.