ROBOT NAVIGATION IN INDOOR ENVIRONMENT THROUGH SELF LEARNING

Indoor robot navigation is a critical functionality for the effective use of mobile robots within indoor environments, especially when GPS signals are unavailable.This paper focuses on addressing the challenges of indoor robot navigation through simulation environments.It utilizes ROS (Robot Operating System), a comprehensive framework comprising libraries for developing and managing robot modules.The simulation is conducted using Gazebo, a tool for creating virtual environments and visualizing data.The process involves collecting environmental data using a LiDAR sensor camera, which captures information about the surroundings.This data is then utilized to train a Machine Learning model, specifically based on Deep Reinforcement Learning techniques.The goal of this model is to plan optimal paths for the robot to navigate within the indoor environment, considering different object settings and layouts.The study evaluates the performance of the Machine Learning model in three distinct environments, each with varying object configurations.This evaluation is crucial as it helps assess the model's adaptability and effectiveness in navigating diverse indoor spaces.The results are presented through score graphs, showcasing how the model's performance varies across different environments and object settings.Overall, this research highlights the importance of leveraging simulation environments, advanced robotics frameworks like ROS, and Machine Learning techniques such as Deep Reinforcement Learning to address indoor robot navigation challenges effectively.