Answers are grounded in this paper and cite their sources.
Develops a data-driven tracking control method in heterogeneous linear systems, suggesting effective control under unknown dynamics.
Optimal tracking control aims to steer a dynamical system along a desired reference trajectory while minimizing a prescribed performance cost. This problem becomes substantially more difficult when both the controlled system dynamics and the target dynamics that generate the reference are not fully accessible. For linear systems, a natural strategy is to first identify the unknown dynamics from data and then solve the resulting optimal control problem, yet this requires identification accuracy sufficient for controller synthesis, which standard methods do not always guarantee. In this paper, we develop a data-driven optimal tracking framework for heterogeneous linear systems with unknown dynamics. An improved dynamic regressor extension and mixing (DREM) scheme is proposed to estimate all unknown system matrices in finite time under mild excitation conditions, providing the exact model knowledge needed for controller design. A discounted optimal control formulation is introduced to handle the parameter mismatch inherent in heterogeneous leader–follower tracking, ensuring the existence and well-posedness of optimal solutions. With the identified model, a gradient-based policy optimization method is employed to compute the optimal tracking controller without requiring a stabilizing initialization. The proposed framework integrates finite-time identification with gradient-based optimization, enabling provably convergent optimal tracking control under unknown dynamics. Simulation results confirm the finite-time parameter convergence of the DREM estimator and demonstrate effective tracking performance under the optimized controller.
No takes yet. Share an insight, caveat, or question.
Mao et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: