Reinforcement Learning Repository of cart pole control system, includes the topics:

► Customisation of the DQN Methodology: Originally developed for Atari game environments, the Deep Q-Network (DQN) was converted for the cart-pole control task.

► High-Dimensional Input: The environment was modified to output video frames of 40x90 resolution

► Agent: Successfully received video frames and Convolution layers were added to the DQN in order to process the high dimension input.

► Comparative Performance Analysis: Despite these enhancements, the convolutional DQN achieved a maximum average reward of 44.9, which is 22.5% of the potential maximum for the Cart-Pole-v1 environment.

► Benchmarking Against Reference Models: Reference models demonstrated superior performance, achieving higher rewards and requiring fewer training episodes to converge.

► Environment adaptation: In this work a system designed to play atari games was modified for solving control task.

► Challenges with High-Dimensional Complexity:The solution for the control task achieved results obtained was lower than simpler methods based on lower dimensional environments.