Embedded Software Implementation of the SISO Adaptive Predictive Control Algorithm

Abstract

Advancements in microelectronic technology provide the opportunity of implementing more complex and computation-heavy algorithms. One area that benefits from these advancements is the automation and control industry. Control techniques and algorithms, previously implemented in industrial and commercial personal computers, are being ported to embedded systems, taking advantage of their real-time and customization capabilities. This work introduces a software-based embedded implementation of a single-input single-output adaptive predictive control (APC) algorithm. The goal of the proposed implementation is to minimize execution time of the APC algorithm without affecting the precision of the results. The ZYBO Zynq-7000 development board was the selected platform for development and evaluation. Our proposed software-centric implementation relies on the development of libraries for matrix data storage and manipulation, and data structures to minimize data transactions. Experimental results included the comparison of APC implementations with different memory usage and data management approaches. An improvement on execution time was possible, reducing it nearly 50% from an initial implementation. Experimental results show no impact on the precision after comparing the implementation results to the results obtained using Scilab-based numerical computation.