Analyze control performance beyond LQR

Background: The study analyzes predictor performance in closed-loop control using Model Predictive Control (MPC) optimized against an infinite-horizon LQR cost, derived from a predictor trained on data.

Question / Future Work: The analysis of predictor performance in closed-loop control needs to be extended beyond the specific LQR setting studied to other control frameworks common in learning-based control, such as model-based reinforcement learning objectives.

Why It Matters: Moving beyond the LQR objective allows the findings to inform model design choices in the broader and more complex domain of policy optimization via model-based reinforcement learning.

Evidence: analyzing predictor performance in closed-loop control in frameworks other than the LQR setting studied in this work, e.g. model based reinforcement learning.