Data-driven inverse optimal control for continuous-time nonlinear systems

Hamed Jabbari

doi:10.1016/j.isatra.2025.12.018

Data-driven inverse optimal control for continuous-time nonlinear systems

2025-12-11

0

PRE

AI

DOI：10.1016/j.isatra.2025.12.018

Original

Original request for help

Save

Abstract

En 中文

• Two new algorithms (model-free and partially model-free) are proposed for inverse optimal control/inverse reinforcement learning in continuous-time nonlinear deterministic systems. • The finite informativity limitation in earlier approaches is addressed, extending applicability to a broader class of input-affine systems. • Computational complexity is reduced by avoiding bi-level optimization; unlike existing model-free meth- ods, the forward problem is solved only during initialization. • Multi-dimensional input systems are supported, including estimation of input-penalty weights, where some existing methods are not applicable. • Stability concerns are mitigated by not applying updated policies to the system at each iteration during estimation.

Data-driven inverse optimal control for continuous-time nonlinear systems

Abstract

Journal

Organization