Research Of Efficiency Improvements Of Predictive Control Based On FPGA

Model predictive control (MPC) is an established control strategy used in the process industry. However, its computational efficiency becomes the main hindrance to its application in higher bandwidth applications, such as in motion control problems. Unlike controllers for process industries, the motion controller must have specific properties, including limited size and high sampling frequency. To meet these requirements, this paper explored the implementation of a new MPC using PSO as the QP-solver, called PSO-MPC, on a field programmable gate array (FPGA) chip. The PSO-MPC-on-FPGA strategy addresses size constraints and satisfies the need for high sampling frequency by exploiting the parallel features of both the PSO-MPC and the FPGA chip.The main contributions are summarized as follows:1. The advantages of FPGA as a new parallel computing platform than traditional structural processors were analysised. At the same time, methods of parallel computing using FPGA on different levels of one algorithm were also given in this paper.2. A new PSO-MPC strategy using PSO algorithm as the QP-solver was proposed. Some changes were made to the standard PSO algorithm to improve its ability of constraints handling and to overcome problems caused by randomness. By the way, the structural flexibility of PSO was the key point for our acceleration of MPC controller.3. A new FPGA programming strategy named'semiautomatic-modular design'was proposed. This new strategy takes the advantages of current available tools for RTL auto-generation, and overcome their insufficient capcity in large scale application.4. A PSO-MPC controller was finally realized on FPGA chip. The goals of volume minimization and acceleration of MPC controller were accomplished.