Implementation And Application Of Nonlinear Model Predictive Controller Based On FPGA

Model predictive control(MPC) is an advanced computer feedback control algorithm which is based on the current state of the system to predict future dynamics. Owing to its series of abilities to handle multi-constraints explicitly, deal with multi-objective optimization problems and have perfect control performance, MPC is widely used in small or metallurgy, pharmaceutical, petrochemical and other slow dynamic systems. While most of the actual systems have nonlinear constraints inherently and work range changing wildly, to the contrary, the control system is becoming increasingly complicated with higher control performance, the MPC has no longer been able to meet the requirements.Therefore, the nonlinear model predictive control(NMPC) has become a hot research topic.NMPC needs to solve the nonlinear programming(NLP) problem with greater computation burden. Online calculation performance has become the bottleneck for the application of NMPC. In order to improve the on-line calculation’s ability of nonlinear model predictive control, this article designs NMPC controller from the perspective of FPGA hardware implementation based on Verilog hardware description language. We adopts particle swarm optimization(PSO)algorithm to solve NLP problem. We finally get high performance NMPC hardware controller considering speed and area through parallel design, assembly line design, resource sharing and so on. To verify the quickness and the effectiveness of the NMPC controller, the trajectory tracking control of wheeled mobile robot’s(WMR) real-time experiment is carried out. Experimental results verify the quickness and the effectiveness of the NMPC controller based on Verilog hardware description language.This paper mainly includes the following three parts:1.Considering the rapid and stable trajectory tracking control requirements and actuators’ constraints of wheeled mobile robot(WMR), establish NMPC controller for trajectory tracking of wheeled mobile robot and complete the off-line verification. In order to meet Matlab toolbox or commercial solver software library is hard to implement onFPGA when solving nonlinear programming(NLP) program, this paper adopts particle swarm optimization algorithm with parallel and simplicity characteristics to solve NLP problem. We establish mathematical model of WMR and code PSO algorithm according to the control requirements to solve NLP problem in MATLAB and then build "NMPC controller- WMR system " closed loop system and complete the WMR trajectory tracking experiment off-line. And finally, we verify the effectiveness of PSO algorithm through comparing with the simulation results solved by "NAG".2.In order to improve the on-line calculation’s ability of NMPC, design NMPC controller based on Verilog hardware description language. Firstly, we design NMPC controller’s top-level circuit structure according to FPGA development process. And then,in order to improve the controller’s speed, we design each module’s micro structures and fixed-point data structures through algorithm analysis and each module can choose different circuit structures and fixed-point data structures. And then, we design objective function module, matrix computation module, random number die block and a triangular function module based on Verilog hardware description language from bottom to top, We determine the optimization solution of each module through trying kinds of architectures,fixed-point data types and function simulation in the Model Sim. And then, we initiate different modules in need and complete the design of NMPC controller through designing state machine for resource sharing and scheduling in the top-level file. Several parallelized identical modules is initiated simultaneously to improve the calculation speed, and control module for same hardware circuit sharing is designed to reduce hardware resource’s consumption. Finally, we design the serial communication module through Verilog description language to realize communication between the NMPC controller and WMR,the structure between the controller and the communication interface based on FPGA is: "receiver module- NMPC controller’s calculation module- transmission module".3.In order to verify the effectiveness and real-time performance of NMPC controller based on FPGA, the hardware in the loop experiment combined with d SPACE for the trajectory tracking of WMR is carried out. Firstly, we build WMR’s trajectory tracking experiment platform based on d SPACE, d SPACE work as real-time simulation system for WMR and FPGA work as the hardware platform for NMPC controller and the two parts form a closed loop system through the serial communication port. We achieve WMR trajectory tracking real-time experiments on the establishing platform and compare computing controllers’ performances among Matlab environment, high level synthesis scheme and Verilog hardware scheme. The calculation time of NMPC controller based Verilog hardware description language is 0.56 ms, and the serial port communication time is1.38 ms, and the wheeled mobile robot sampling time is 10 ms. The experimental resultsverify the real-time and effectiveness of NMPC controller based on FPGA chip.This article designs a nonlinear model predictive controller through Verilog hardware description language based on FPGA. We give detail design procedures and methods include of the offline controller design, the choice of fixed-point data types, the designs of each modules’ micro architecture and real-time experiment which have significance. But there are still some problems needed to be solved:(1)Convergence of nonlinear model predictive control algorithm and constraint particle swarm optimization algorithm remain to be considered. This paper adopts third order wheeled mobile robot system to verify the effectiveness of the nonlinear model predictive controller based on FPGA, we have not considered the convergence stability problem of the system. We will solve the convergence problem in theory by discussing terminal domain in order to apply to more complex nonlinear dynamic system quickly.(2)We will carry out physical control experiment for wheeled mobile robot. This paper carries out hardware in the loop experiment to wheeled Mobile Robot with d SPACE simulation system, The physical control experiment on laboratory’s P3 DX wheeled Mobile Robot physical platform need to be done.