中文摘要：

针对多变量连续时间广义预测控制提出了一种新的时滞解决方案及其参数递推计算方法（MDCGPC）．在多变量子子模型的基础上采用时滞预测机制，通过参数恒等式变换，将时滞补偿环节引入到预测输出中，通过求解二次性能指标获得控制律的显式解．避免了模型时滞近似方法带来的误差和非最小相位特性，可以处理各路时滞不相同的情况，通过对参考轨迹的调节，可以具有一定的解耦能力．同时参数的递推计算方法简化了多项式数学运算的难度，减小了计算负荷和复杂性．仿真结果验证了所提出算法的有效性．

英文摘要：

A new delay predictive solution and the parameters recursive computation method are proposed for the multivariable continuous-time generalized predictive control （MDCGPC）. By using the delay predictive mechanism and the computation of the identities, the delay compensator is introduced in the predictive output based on the multivariable subsubmodels, and the explicit control law is obtained by solving the QP cost function. This control strategy can avoid the model error and nonminimum phase characteristic caused by the time-delay approximation method. It can solve the system with multiple timedelays and obtain decoupling performance by tuning the reference trajectory. Moreover, the recursive computation of control parameters reduces the computation load and complexity. Some simulations verify the performance.