Smart Predict-Then-Control: Control-Aware Surrogate Refinement for System Identification
This addresses the problem of improving control performance in dynamic systems like quadrotors, though it appears incremental as it refines existing prediction-oriented models.
The paper tackles the problem of model-based control by introducing Smart Predict Then Control (SPC), a control-aware refinement procedure that optimizes surrogate objectives to improve prediction models based on induced control actions. On a wind-disturbed quadrotor trajectory tracking task, SPC reduces tracking RMSE by 70% and closed-loop cost by 42% compared to a baseline.
This paper introduces Smart Predict Then Control (SPC), a control aware refinement procedure for model based control. SPC refines a prediction oriented model by optimizing a surrogate objective that evaluates candidate models through the control actions they induce. For a fixed surrogate variant under unconstrained control, we establish the smoothness of the surrogate, projected gradient convergence at a sublinear rate of order one over K, and a bias decomposition that yields a conditional transfer diagnostic. On a wind disturbed quadrotor trajectory tracking task, Updated SPC reduces tracking RMSE by 70 percent and closed loop cost by 42 percent relative to the nominal baseline.