Abstract— This paper presents a novel, computationally feasible procedure for computing optimal switching surfaces, i.e. optimal feedback controllers for switched autonomous nonlinear systems. The systems under consideration are regulated by appropriately scheduling their operation modes. Given a finite mode sequence, the control task requires to determine switching surfaces, which implicitly encode the optimal switching times for a family of trajectories emerging from a prespecified initial state set. Optimality of the switching times is assessed according to a nonlinear performance criterion.