In the paper:
TøNdel, P., Johansen, T. A., & Bemporad, A. (2003). An algorithm for multi-parametric quadratic programming and explicit MPC solutions. Automatica, 39(3), 489-497.
I have seen the following part which indicates that the terminal constraint in Model Predictive Control (MPC) could lead to degeneracy. However, I am not convinced how this happens?
In an MPC problem one might avoid full-dimensional critical regions with violation of LICQ by simply slightly perturbing the weight matrices and the constraints, without producing significant changes of the closed-loop behavior. On the other hand, in some situations this may not be possible, for instance equality constraints such as terminal state constraints $x_{t+N|t}=0$ , would lead to violation of LICQ (cf. Berkelaar et al., 1997, Example 6.3). In such cases, full-dimensional critical regions can be handled by solving a QP, as in Tondel, Johansen, and Bemporad (2001b).
LICQ= linear independence constraint qualification
It refers to the following paper which I cannot find on the Internet (There is a version which looks different as it does not have Section 6.3).
Berkelaar, A. B., Roos, K., & Terlaky, T. (1997). The optimal set and optimal partition approach. International series in operations research and management science, 6-1.
Would anyone please explain how a zero terminal constraint leads to the redundancy of constraints?