推导了多变量广义预测控制算法，该算法能有效地处理过程的不稳定性和非最小相位特性，并具有良好的解耦性能。同时提出一种处理纯滞后对象的新方法，即将近似法和广义预测控制相结合解决时滞的方案。这种方法可以解决通道之间时滞不同的对象，并且针对广义预测控制算法的复杂性对其简化算法（多变量单值广义预测控制算法）进行了研究。通过仿真研究表明，多变量单值广义预测控制算法在跟踪性、鲁棒性、抗干扰能力等方面都得到很好的控制效果。

Multivariable generalized predictive control (MGPC) algorithm was deduced in this paper. It can effectively handle the instability and the non-minimum phase of the process. Furthermore, it has good decoupling performance. A new method to handle the pure delay of a process by combining the approximation method with the generalized predictive control was roposed, and the different time delay could be handled. A reduced method was presented i.e. multivariable single generalized redictive control algorithm to simplify the complexity of MGPC. Simulation shows that this algorithm achieves a good result in tracing property, robustness, anti-interference etc.