Certified exact error-margin evaluation and input-tightening analysis for Tube MPC
Develop a certified truncation bound for evaluating the exact worst-case tracking-error margin from the convergent matrix-power series, and quantify the severity of the input-tightening reservation required for the ancillary correction in Tube MPC under bounded actuation, including its growth with the sampling period.
References
With a certified bound on the truncation of the series, this exact value can be evaluated more cheaply than the DARE solve it depends on, which removes the practical motivation for using an approximate formula at all. The tube construction also imposes a second tightening, on the input rather than the corridor, not discussed here: the ancillary correction $-Ke_k$ must itself be reserved out of $u_{\max}$, and this reservation can be severe in practice, growing sharply with the sampling period. Both points are left for future work.