摘要：AbstractWe present a computationally simple strategy to solve model predictive control problems with mixed-integer controls. We restrict the set of possible discrete control sequences considered at each time step by using parameterized quasi-translations, which relate discrete controls over consecutive timesteps. The parameters can be tuned to balance controller performance, chatter and turnaround time. Using quasi-translations we design two NMPCs for autonomous vehicle control with discrete gear input and exclusively acting pedal inputs. We compare these to an NMPC using a representative ‘out-of-the-box’ MINLP solver. The quasi-translation controllers run orders of magnitude faster with minimal degradation in tracking performance. This is an exciting development for integer controls that begs future work.
