Gridworld & CartPole Solver

This project implements and compares diffeernt Reinforcement Learning approaches to solve the Gridworld and CartPole problems.

Markov Desicion Proccess (MDP)

Use of an MDP structure + Value Iteration Algorithm to solve the Gridworld board problem

Q-Learning

Use of Q-Learning + Value Iteration Algorithm to solve the Gridworld board problem

Deep Q Network

Implemented 4 different cases to solve the CartPole problem

• Conventional Deep Q Network
• Deep Q Network with Experience replay and Target Netwrok
• Deep Q Network with Target Netwrok
• Deep Q Network with Experience replay and Target Netwrok (best accuracy level)