Dynamic Programming (Approximate, Differential), Model Predictive Control

[Kat007]

Hello,

Could someone please explain in a short summary what the difference between the following are:

1. Dynamic Programming from Control Engineering literature and Dynamic programming from Reinforcement learning literature?
2. Approximate Dynamic Programming vs Differential Dynamic Programming vs Dynamic Programming itself?
3. Dynamic Programming and Linear Quadratic Regulator?
4. Linear Model Predictive Control and Nonlinear Model Predictive Contro? (Specifically, this technique requires full knowledge of the dynamics right? So what exactly does it predict n steps ahead?)

I am very familiar with Linear Quadratic Regulator and Dynamic Programming coming from the Reinforcement Learning literature.

Please let me know if you could help.

Thank you very much

 

[Navin A S]

.1. Dynamic Programming from Control Engineering literature: This is a technique used in control engineering to solve complex optimization problems. It involves breaking down an optimization problem into simpler subproblems that can be solved iteratively, and then combining these solutions to find the best overall solution.2. Approximate Dynamic Programming vs Differential Dynamic Programming vs Dynamic Programming itself: Approximate Dynamic Programming is a method of solving dynamic programming problems using approximate solutions, while Differential Dynamic Programming is a numerical optimization technique for solving optimal control problems. Dynamic Programming itself is a general optimization technique for solving complex problems by breaking them down into smaller, simpler subproblems.3. Dynamic Programming and Linear Quadratic Regulator: Dynamic Programming is a general optimization technique, while a Linear Quadratic Regulator is a specific type of controller used to regulate a system's behavior given a set of objectives.4. Linear Model Predictive Control and Nonlinear Model Predictive Control: Linear Model Predictive Control (MPC) is a technique that uses linear models to predict future states of a system, while Nonlinear MPC uses nonlinear models to do the same. The technique requires full knowledge of the system's dynamics in order to make accurate predictions, and can be used to predict multiple steps ahead.