AUTONOMOUS NAVIGATION USING REINFORCEMENT LEARNING

ABSTRACT

This project investigates the application of reinforcement learning (RL) for autonomous drone navigation in unseen environments where the destination of the drone changes over a period of time. We propose a system where a drone learns to navigate through trial and error, receiving rewards for desired behaviors and penalties for undesired actions. This approach allows the drone to adapt to complex situations without pre-programmed flight paths. The drone acts as an agent interacting with its environment through sensors. Sensor data provides information about the state (e.g., position, obstacles). Based on the state, the drone takes actions (e.g., move forward, yaw rotation). The environment provides rewards for reaching the goal location and penalties for collisions or straying from the designated area. Through repeated interactions, the RL algorithm learns an optimal policy for navigating the environment.