| Model predictive control(MPC) is an optimization algorithm based on online computation, and it has the ability to handle multi-parameter optimization control problem with constraints uniformly. But MPC needs to choose larger predictive domain and control domain when the structure and environment of the controlled object get more complex, which leads to increasing the online computation time and reducing the control effect. From the current algorithm, MPC can usually only be used in the systems with large sampling time and slow dynamic. It has great theoretical and applied value to research a fast model predictive control algorithm.In order to reduce MPC online computation complexity and online computation time, two kinds of MPC which based on CVXGEN and online optimization respectively are introduced, also the two algorithms’ principle and performance are analyzed. Finally, the two algorithms are applied to the three-DOF helicopter system, and compared with the traditional MPC and the explicit MPC.In this paper, the main work and achievements are as follows:(1) The research status and developments of the MPC, fast MPC and control methods of three-DOF helicopter are summarized. The basic principle of MPC and convex optimization theory are introduced, all of which provide the theoretical basis for the following study.(2) Because the large and medium problem explicit MPC can’t be handled, the fast MPC algorithm based on CVXGEN is proposed. First the basic principle of the software tool CVXGEN is introduced, and the fast solver generated by CVXGEN which are combined with MPC can be used to solve a kind of MPC problems. The algorithm is applied to numerical examples, also the performance and its fast ability to solve problems are analyzed.(3) Because not all problems fast MPC algorithm based on CVXGEN can be handled, the fast online model predictive control is proposed. The algorithm exploits improved interior point method and warm-start to reduce online computation complexity and time. Finally the algorithm is applied to large, medium and small sized numerical examples, the ability to handle all kinds of problems quickly is verified.(4) Finally, the above two algorithms are applied to the three-DOF helicopter model, and compared with MPC and explicit MPC under the same condition. The control performance, online computation time and error etc. of these four algorithms are compared and analyzed, also the advantages and disadvantages are summarized.(5) Summary of this paper is given, and the future research direction is discussed. |
