Cartpole Balancer Using Deep Q-Network

In this paper, we purely focused on the algorithm called Deep Q-Network (DQN) to solve CartPole environment, a classic reinforcement learning task. The goal of the Cartpole task is to balance the pole on a moving cart by applying the sequences of left or right forces. In traditional approaches, such as Manual Rule-Based Controller, Random Agent Method and many are there which are struggling to deal with the complexity of this task. In contrast, DQN leverages deep learning to approximate the optimal action-value function, enabling the agent to learn a policy that maximizes cumulative reward through interactions with environment. This project implements DQN algorithm, training a deep neural network to predict Q-values for the different state-action pairs. Through iterative updates and experience replay, the agent learns to maintain the pole in an upright position for extended periods. The performance of the DQN agent is evaluated based on its ability to achieve and maintain balance, demonstrating the effectiveness of deep reinforcement learning in complex control tasks.