无记忆自适应子空间预测控制器设计

Memoryless Self-adaptive Subspace Predictive Controller

针对遗忘因子作用于过去与将来Hankel矩阵权重不同的特点,文中提出一种无记忆自适应子空间预测控制器设计方法。该方法利用当前变遗忘因子构造了过去与将来Hankel矩阵的无偏更新形式,并根据固定区间内期望指令与预测输出的误差值实现了变遗忘因子的自适应更新,从而使算法既避免了因频繁更新遗忘因子导致辨识精度变差的问题,又提高了算法的控制性能。仿真算例验证了算法的有效性。

A new memoryless self-adaptive subspace predictive controller was proposed for different weight for the past and future of Hankel matrices under the current forgetting factor.Based on the upper theory,the new unbiased update form of Hankel matrices was designed with the current factor.In order to avoid incidental disturbance of noise existing in traditional algorithm caused by frequent factor update,the adaptive updating form of variable forgetting factor was designed on the distance value of expectancy instruction and predictive output at fixed interval.The modified predictive controller has better performance.Its efficiency was validated by a simulation example.