Adaptive Horizon Model Predictive Control and Regulation, Short Horizon Estimation
Abstract
We shall develop Model Predictive Feedback (MPF). Instead of waiting until the next state is known to compute the appropriate value of the control, MPF computes a localfeedback around the predicted value of the next state and when the next state becomesknown it uses the feedback. Since the computation of the feedback can be started beforethe next state is known, we have more time to compute it, so MPF can be used on fasterprocesses. And having a feedback around the nominal optimal trajectory at the current timeadds robustness and may allow us to avoid going to the solver for one or more time steps.We shall also develop Adaptive Horizon Model Predictive Regulation (AHMPR), a technique for tracking a reference using MPC techniques. Like AHMPC, AHMPR will verify inreal time that asymptotic tracking is being achieved.The state estimation problem is dual to the state stabilization problem. Moving HorizonEstimation (MHE) use MPC techniques to estimate the state from partial and noisy measurements. Its roots go back to Minimum Energy Estimation (MEE). We propose to developa discrete time MEE method that will allow us to simplify MHE by making the horizon asshort as possible, one time step. We will call this Short Horizon Estimation (SHE).In summary we offer a suite of techniques to speed up Model Predictive Control andMoving Horizon Estimation so that they can be used on fast, nonlinear processes such asaircraft.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jul 28, 2017
- Source ID
- FA95501710219
Entities
People
- Arthur J. Krener
Organizations
- Air Force Office of Scientific Research
- United States Air Force
- University of California, Davis