A New Constrained Cost Value Iteration for Optimal Control of Discrete-Time Nonlinear Systems

Tao Li,Qinglai Wei
DOI: https://doi.org/10.1109/cac53003.2021.9727847
2021-01-01
Abstract:In this paper, a new constrained cost value iteration (CCVI) adaptive dynamic programming (ADP) algorithm is developed to solve optimal control problems with constrained cost function. The CCVI algorithm is initialized by a feasible control law, and the cost function of this initial feasible control constitutes the constraint condition in the policy update step. The convergence and optimality of the CCVI are analyzed. Furthermore, the feasibility of any iterative control law and the on-line evolving control law are proven. It is emphasized that the feasibility of any iterative control law and the on-line evolving control law guarantee that the CCVI algorithm can be used in situations where the cost function constraint is extremely important. Finally, a simulation example is given to verify the effect of the developed algorithm.
What problem does this paper attempt to address?