General Smith Predictors from an Observer-Controller perspective

The Smith predictor is a well-known method for designing controllers for plants with I/O delays. Assuming the time delay known, this control configuration compensates the delay within the loop making it possible to design the controller just considering the (rational) delay-free part of the plant. Since the original Smith predictor does not allow to deal with unstable plants, intensive research has been conducted on the control of unstable delayed processes during the last years, giving rise to modifications and alternative structures. This work addresses the control of SISO unstable delayed systems through a generalized Smith predictor based on the so called observer-controller configuration. Once the continuous-time derivation is carried out, the discrete time version of the scheme is considered attending to implementation issues. The presented scheme introduces a general framework where many of the recently proposed dead-time compensators fit in as particular cases, providing thus more insight into the problem.