“…The finite-time optimal controller has to account for possible switches and the set of initial states is generally non-convex, making it costly to find explicit offline solutions [7], [8], [14], [23]. Guaranteeing recursive feasibility and Lyapunov stability in the receding-horizon implementation requires a terminal invariant set and a CLF on that set [13], [24], [25]. If the closed-loop equilibrium is a shared point of some regions, finding such a CLF is non-trivial and needs the methods of [4], [5], [9], [12], [22], bringing once again the issues of the S-procedure, and initialization, and respecting constraints.…”