Inverse Continuous-Time Linear Quadratic Regulator: From Control Cost Matrix to Entire Cost Reconstruction
Yuexin Cao, Yibei Li, Zhuo Zou, Xiaoming Hu
TL;DR
The paper addresses inverse optimal control for the continuous-time finite-horizon LQR, aiming to identify the cost matrices $R$, $Q$, and $F$ from observed optimal inputs. It introduces two $R$ reconstruction strategies—one using the full trajectory of the optimal feedback $K(t)$ and a second using samples at selected times—providing necessary and sufficient conditions for uniqueness and explicit solution characterizations, including the impact of controllability on well-posedness. For $Q$ and $F$, it studies identifiability under a given $R$, deriving analytic recovery formulas when controllability holds, and LMIs-based descriptions of solution spaces otherwise; it also analyzes the bijection between $Q$ and $F$ and presents explicit expressions when one is fixed. A structural equivalence is established between the inverse problems with and without known $R$ via an auxiliary LQR problem, enabling reformulations that translate time-varying IOC to a (potentially) time-invariant LQR setting. Collectively, these results advance practical IOC for continuous-time systems, with implications for real-time identification and cross-domain applicability.
Abstract
This paper studies the inverse optimal control problem for continuous-time linear quadratic regulators over finite-time horizon, aiming to reconstruct the control, state, and terminal cost matrices in the objective function from observed optimal inputs. Previous studies have mainly explored the recovery of state cost matrices under the assumptions that the system is controllable and the control cost matrix is given. Motivated by various applications in which the control cost matrix is unknown and needs to be identified, we present two reconstruction methods. The first exploits the full trajectory of the feedback matrix and establishes the necessary and sufficient condition for unique recovery. To further reduce the computational complexity, the second method utilizes the feedback matrix at some time points, where sufficient conditions for uniqueness are provided. Moreover, we study the recovery of the state and terminal cost matrices in a more general manner. Unlike prior works that assume controllability, we analyse its impact on well-posedness, and derive analytical expressions for unknown matrices for both controllable and uncontrollable cases. Finally, we characterize the structural relation between the inverse problems with the control cost matrix either to be reconstructed or given as a prior.