Abstract
Modern data-driven control applications call for flexible nonlinear modelsthat are amenable to principled controller synthesis and realtime feedback.Many nonlinear dynamical systems of interest are control affine. We propose twonovel classes of nonlinear feature representations which capture control affinestructure while allowing for arbitrary complexity in the state dependence. Ourmethods make use of random features (RF) approximations, inheriting theexpressiveness of kernel methods at a lower computational cost. We formalizethe representational capabilities of our methods by showing their relationshipto the Affine Dot Product (ADP) kernel proposed by Casta\~neda et al. (2021)and a novel Affine Dense (AD) kernel that we introduce. We further illustratethe utility by presenting a case study of data-driven optimization-basedcontrol using control certificate functions (CCF). Simulation experiments on adouble pendulum empirically demonstrate the advantages of our methods.