摘要：AbstractAccurate models of inverse systems are required for high performance in inverse model-based feedforward control. Identification of inverse systems can be challenging, especially if the inverse system has poles outside the typical stability region. The aim of this paper is to estimate non-causal models of inverse systems, for intended use in feedforward control, where non-causality can be exploited to compensate ‘unstable’ poles. The developed method employs kernel-based regularization to improve the bias/variance trade-off, where the non-causal kernel is constructed using rational basis functions that include poles outside the usual stability region. The benefits of the developed method are demonstrated on an example, including non-causality.
