On the design of a neural network chip for control applications

The design of a neural network chip is presented to be used in control applications. Specifically, the chip will be integrated in the control of a commercial DC motor having a significant amount of static friction. The main functions of the chip are to identify the friction and then to produce the exact amount of torque required to compensate for that friction. The author describes the neural network model and its hardware design, implementation, and simulation results.<
>