Reinforcement learning Algorithms such as SARSA, Q learning, Actor-Critic Policy Gradient and Value Function Approximation were applied to stabilize an inverted pendulum system and achieve optimal control. So essentially, the concept of Reinforcement Learning Controllers has been established. The Reinforcement Learning Controllers have been compared on the basis of performance and efficiency and they are separately compared with the classical Linear Quadratic Regulator Controller. Each of the RL controller have been integrated with a Swing up controller. A virtual switch toggles between the Swing up controller and the RL controller automatically, based on the value of the angular deviation theta with respect to the vertical plane. My research paper and my undergraduate thesis have been uploaded for reference. All the codes have also been uploaded.
Dynamic Programming was applied to Jack's Car Rental Problem to maximize the amount Jack gets by the end of the day, with some Constraints.
imagef/Reinforcement-learning-Algorithms-and-Dynamic-Programming
Reinforcement learning Algorithms such as SARSA, Q learning, Actor-Critic Policy Gradient and Value Function Approximation were applied to stabilize an inverted pendulum system and achieve optimal control. So essentially, the concept of Reinforcement Learning Controllers has been established. The Reinforcement Learning Controllers have been compared on the basis of performance and efficiency and they are separately compared with the classical Linear Quadratic Regulator Controller. Each of the RL controller have been integrated with a Swing up controller. A virtual switch toggles between the Swing up controller and the RL controller automatically, based on the value of the angular deviation theta with respect to the vertical plane. My research paper and my undergraduate thesis have been uploaded for reference. All the codes have also been uploaded.
MATLAB