Page 158 - Robots Androids and Animatrons : 12 Incredible Projects You Can Build
P. 158

If you want to experiment, begin by using the pulsout and pwm
                                       commands to vary the DC motor speed.
                                       Neural sensors (logic)
                                       With a small amount of programming, we can change our fuzzy logic
                                       sensors (CdS photocells) to neural sensors. Neural networks are an
                                       expansive topic. We will limit ourselves to one small example. For
                                       those who want to further pursue study into neural networks, I
                                       recommend  a  book  I’ve  written  titled  Understanding  Neural
                                       Networks (Prompt, Indianapolis, 1998, ISBN 0-7906-1115-5).

                                       To create neural sensors, we will take the numeric resistive reading
                                       from each sensor, multiply it by a weight factor, and then sum the
                                       results. The results are then compared to a tri-level threshold value
                                       (see Fig. 6.36).
                                       Thus our small program and sensors are performing all the func-
                                       tions expected in a neural network. We may even be pioneering a
                                       neural first, by applying a multivalue threshold scheme. Do multi-
                                       value thresholds exist in nature (biological systems)? The answer
                                       is yes. For instance, an itch is an extremely low level of pain, and
                                       the sensation of burning can be felt when sensing something ice
                                       cold or hot.                                                       137






























                                         6.36 Schematic of simple tri-level neuron


                                                       Team LRN                                      Intelligence
   153   154   155   156   157   158   159   160   161   162   163