噪声未知的对偶模型预测控制算法研究

Research on Dual Model Predictive Control Algorithm with Unknown Noise

模型预测控制(model predictive control,MPC)是处理实际系统中存在的各种约束条件的最有效方法之一。针对具有未知噪声的随机系统,提出了一种具有对偶学习特点的模型预测控制算法。使用Kalman滤波器对系统状态进行估计,由估计得到的当前时刻的状态作为预测系统未来动态的起点,系统噪声的结构特性已知,其方差在有限集合里,使用后验概率,在优化控制目标的同时学习出噪声未知方差的真值。仿真结果表明,相较于传统的模型预测控制算法,在具有不确定性的系统中,对偶模型预测控制算法在驱动控制系统朝着期望方向运行的同时,还能够对未知参数进行有效地学习。

Model predictive control(MPC) is one of the most effective methods to deal with various constraints existing in the actual system.For the stochastic system with unknown noise,a model predictive control algorithm with dual learning characteristics is proposed.Kalman filter is used to estimate the system state.The state of the estimated current time is the starting point of the future dynamic prediction system.The structural characteristics of the system noise are known.The variance of the system noise is in a finite set,and the posterior probability is used to learn the true value of unknown variance of noise while optimizing the control target.Compared with the traditional model predictive control algorithm,in the uncertain system,the simulation results show that the dual model predictive control algorithm can drive the control system to run in the desired direction,but also can effectively learn the unknown parameters.