Convergence in unconstrained discrete-time differential dynamic programming

摘要: The conditions under which the original differential dynamic programming (DDP) algorithm can be expected to converge are investigated and modifications in improve convergence properties proposed. Quadratic of DDP requires that stagewise Hessian matrices computed positive definite. Three procedures for modifying guarantee situations not definite One (an active shift) converges quadratically. computational requirements with without shift presented, impact ratio state variable dimension control on relative efficiency different is discussed. Numerical results a series problems, including one 16 variables 200 time steps, presented. most robust procedure an adaptive utilizes constant followed by shift. >