Adaptive Dynamic Programming and Reinforcement Learning (ADPRL) for nonlinear optimal control

Adaptive dynamic programming (ADP), also known as approximate dynamic programming, neuro-dynamic programming, and reinforcement learning (RL), is a class of promising techniques to solve the problems of optimal control for discrete-time (DT) and continuous-time (CT) nonlinear systems.

MATLAB codes of ADPRL, including iterative and online ADPRL, are provided. Currently added ADPRL algorithms include:

• Value iteraion for DT systems
• Value iteraion for DT systems (positive semi definite initial value function)
• Policy iteration for DT systems
• Policy iteration for CT systems
• Integral reinforcement learning for partially unknown CT systems
• Model-free integral reinforcement learning for completely unknown nonaffine CT systems
• Model-free nonzero-sum games for completely unknown nonaffine CT systems
• Online learning policy update for DT systems
• Online learning without initial admissible control for CT systems
• Parallel control-based optimal tracking for nonaffine CT systems

Requirements

---

• MATLAB