Does Optimal Control Always Benefit from Better Prediction? An Analysis Framework for Predictive Optimal Control

TL;DR

This paper introduces a new analysis framework for predictive optimal control that accounts for imperfect predictions, proposing the hidden prediction state concept and emphasizing control performance-based predictor evaluation.

Contribution

It develops a novel framework incorporating subjective beliefs and predictor evaluation methods, highlighting the importance of control performance metrics over traditional error measures.

Findings

Mean squared error and likelihood do not guarantee control cost improvement.

Control performance-based evaluation better correlates predictor quality with control outcomes.

Numerical and real-world automotive examples validate the framework.

Abstract

The ``prediction + optimal control'' scheme has shown good performance in many applications of automotive, traffic, robot, and building control. In practice, the prediction results are simply considered correct in the optimal control design process. However, in reality, these predictions may never be perfect. Under a conventional stochastic optimal control formulation, it is difficult to answer questions like ``what if the predictions are wrong''. This paper presents an analysis framework for predictive optimal control where the subjective belief about the future is no longer considered perfect. A novel concept called the hidden prediction state is proposed to establish connections among the predictors, the subjective beliefs, the control policies and the objective control performance. Based on this framework, the predictor evaluation problem is analyzed. Three commonly-used predictor evaluation measures, including the mean squared error, the regret and the log-likelihood, are considered. It is shown that neither using the mean square error nor using the likelihood can guarantee a monotonic relationship between the predictor error and the optimal control cost. To guarantee control cost improvement, it is suggested the predictor should be evaluated with the control performance, e.g., using the optimal control cost or the regret to evaluate predictors. Numerical examples and examples from automotive applications with real-world driving data are provided to illustrate the ideas and the results.