Fuzzy logic and neural net control for the “Smarter Car”

As a requirement of the course Laboratory Introduction to Embedded Control (LITEC)-a junior-level general engineering course at Rensselaer Polytechnic Institute-the students build a “Smart Car” that follows a racetrack laid out on the floor. This paper discusses the implementation of the “Smarter Car” that relies on low-resolution camera images of the racetrack for training a neural net off-line for car position and road curvature. A fuzzy controller determines the required steering action to keep the car on the track. This paper illustrates how neural nets that were software trained can be implemented in an off-the-shelf Motorola 68HC11 embedded microprocessor system. Even though just one specific example is illustrated in detail, this approach is generic and can be readily extended for the implementation of similar systems such as the control of robot arms and more general industrial neural net and fuzzy logic applications