CMU CAM :

BOB 3 has now a CMUCAM on board. This camera can detect an object that has a predefined color, in an image. The camera is intefaced to the FoxBoard via a serial connexion. This camera is very easy to use; it computes itself the image processing needed.

CMU CAM

Ultra Sonic range finder:

The US sensors are SRF04 (Devantech); They give a measure of distance to obstacle, between 4cm and 2m. The detection cone is quite large: more than 35deg.

A large cone is usefull to detect obstacles. Obstacles are detected quite systematically. But those sensor are quite slow. I only reach 8 measurement per second. This is not enough. The interpretation of the measurement is directly made by the PicBasic, which measures the time of the pulse out of the sonar. I use only one input of the PicBasic (connected to the 5 sonars via diodes), and 5 outputs for controling each sonar.

I don't use Infra red range finders, because they have a too narrow cone of detection, and cannot detect all the obstacles.

Sonar SRF04

---

Wheel encoders:

The optical wheel encoders are the most important sensors of the robot. They give the robot a good estimation of its position. They measyre the progress of each wheel. Via quadrature encoders, we can determine the direction of rotation.

The wheels are made with wheels of ball mouse. Those sensors very simple, but robust, not sensible to ambiant light, and alignment errors.

The encoder wheels are directly mounted on robot wheel axes. The small gearbox has a lot of backlash. In this situation, the encoder must be placed closed to the wheel, and not to the motor.
The optical sensors can be ajusted in 3 directions. This adjusment is essential for a good behaviour. The value of the resistors can also be changed so that the sensor be more robust. I decided to increase the current in the emitter diode so that the sensor be less sensitive to ambiant light.

The information given from those encoder is essential. It is an estimation of the coordinates of the robot. The robot computes those coordinates 40 times per second. The overall accuracy is quite good, compared to the simplicity of the sensors.