3D Reconstruction in Robotics: A Comprehensive Review

Abstract

In this paper, we delve into the swiftly progressing domain of 3D reconstruction within robotics, a field of critical importance in contemporary research. The inherent potential of robots to engage with and understand their environment is significantly enhanced by integrating 3D reconstruction techniques, which draw inspiration from the complex processes of natural evolution and human perception. This study not only highlights the importance of 3D reconstruction methodologies in the broader context of technological advancement but also outlines their pivotal contributions to the field of robotics. Humans have evolved over millions of years to adapt to their surroundings through natural selection, enabling them to perceive the world. 3D reconstruction methods are inspired by natural processes to replicate objects, providing more detailed information about the perceived object. With this approach to object perception, robotics plays a crucial role in utilising these techniques to interact with the real world. Our study illustrates recent advancements in applying 3D reconstruction methods within robotics and discusses necessary improvements and applications for future research in the field. • Research status of Neural Radiance field, Simultaneous Localisation and Mapping (SLAM), Semantic, Amodal 3D Reconstruction, Gaussian splatting, and NeRF in robotics were reviewed. • Qualitative and Quantitative comparisons on prominent 3D reconstruction techniques were analyzed. • Datasets used for prominent 3D reconstruction methods in robotics were summarized.