Enhancing Robot Self-Localization Accuracy and Robustness: A Variational Autoencoder-Based Approach

Abstract

Self-localization is a crucial task for robots, demanding high accuracy. In this work, we propose a new robot localization method based on the Variational Autoencoder (VAE). In our method, the robot utilizes the captured image to generate robot localization in indoor environments. The utilization of VAE makes the system adaptive to varying environmental conditions. Our findings demonstrate that utilizing both the robot's coordinates and images as training data significantly enhances the accuracy of robot self-localization estimation and improves the robustness of the system due to sensor data noise.