Linear model predictive control (MPC) can track reference trajectories better than LQR control by considering future states over a finite time horizon at each time step. MPC formulates an optimization problem to minimize a quadratic cost function subject to model dynamics and input/state constraints, to compute a non-stationary control policy. This is in contrast to LQR which computes a stationary policy by optimizing over an infinite horizon. MPC allows the controller to "predict" future states using a dynamical model and minimize errors between predicted and reference states/inputs.