Abstract
We present a new method for learning control law that stabilizes an unknownnonlinear dynamical system at an equilibrium point. We formulate a systemidentification task in a self-supervised learning setting that jointly learns acontroller and corresponding stable closed-loop dynamics hypothesis. Theopen-loop input-output behavior of the underlying dynamical system is used asthe supervising signal to train the neural network-based system model andcontroller. The method relies on the Lyapunov stability theory to generate astable closed-loop dynamics hypothesis and corresponding control law. Wedemonstrate our method on various nonlinear control problems such as n-Linkpendulum balancing, pendulum on cart balancing, and wheeled vehicle pathfollowing.