These codes produce the results of our proposed algorithm, shown in the paper https://arxiv.org/abs/2103.12351. The Region of Attraction (ROA) of our controller (yellow set) contains about 98% of the ROA of Tube MPC (grey set, https://www.sciencedirect.com/science/article/pii/S0005109803002838?via%3Dihub, Section 5). The yellow ROA is about 4% larger in volume.
The following table of computation times demonstrates that our approach is up to 15x faster at online control computations compared to the aforementioned Tube MPC.
Additionally, even with an open-loop policy sequence (i.e., without re-solving the MPC problem), we obtain a safe set (yellow set), which is about 12x larger in volume than the ROA of the System Level Synthesis based Constrained LQR (blue set, https://arxiv.org/pdf/1809.10121.pdf, Section 2.3).
