A LOW-POWER APPLICATION-SPECIFIC INSTRUCTION PROCESSOR FOR MODEL PREDICTIVE CONTROL APPLICATIONS

摘要

This paper presents a low-power microcontroller for embedded real-time model predictive control applications. High speed and low power consumption in the processor are achieved by reducing the precision in the data representation, coupled with using logarithmic number system arithmetic instead of the classical IEEE 754-like floating point. Custom architecture details are presented and with the use of finite element method simulations the overall performance of the controller is provided. In order to compare the performance with that of alternative control approaches a proportional integral controller is also examined. With this work we propose a practical solution to a very complex problem: embedding model predictive control.