Q-Learning
An End-to-End Zero Shot Framework for Autonomous Drone Navigation
Members: Arnav Kharbanda*, Yashasav Prajapati*, Gopal Bansal*, Arpit Gautam*
I reimplemented the Q-Learning algorithm, a foundational reinforcement learning technique, to gain a deeper understanding of its mechanics and practical applications.
- Objective: To train an agent to learn optimal policies through trial and error by maximizing cumulative rewards in a given environment.
- Approach: Implemented the algorithm from scratch, focusing on the Q-value update rule and exploring various learning rates and discount factors to optimize convergence.
- Outcome: Successfully tested the implementation in grid-world scenarios and observed the agent improving its performance over episodes, effectively navigating towards the goal state.
This exercise enhanced my knowledge of reinforcement learning concepts and strengthened my ability to build algorithms from first principles.
