Gaussian Splatting for Enhanced Scene Representation in 5G-Enabled Virtual Reality for Remote Operations

Scene reconstruction in virtual reality (VR) applications presents numerous challenges, particularly in scenarios requiring remote communication, such as with remotely operated vehicles (ROVs). As a result, there have been a number of proposed solutions for addressing these concerns more effectively and efficiently. Most of these techniques for representing 3D environments often struggle with training and rendering speed, accuracy, and the ability to handle complex, dynamic environments, which limits their effectiveness in time-sensitive fields which require remote operations. This thesis focuses on the application of the 3D Gaussian splatting technique for real-time radiance field rendering aiming to improve the efficiency and fidelity of 3D scene representation for remote operations. The ROV is equipped with high-resolution cameras to capture detailed visual data, facilitating faster and more accurate reconstruction of remote environments. Gaussian splatting allows for smoother transitions in scene representation, making it ideal for real-time applications where both speed and visual quality are crucial. This work explores the deployment of Gaussian splatting to enhance scene reconstruction in remote settings. Additionally, a framework for integrating Gaussian splatting with 5G technology, leveraging high-speed, low-latency communication to improve overall operational effectiveness and user experience in remote and critical scenarios, is proposed.