CEDoc-UM6P-AI MOV : Visual Simulteneous Localization and Mapping for Mobile Full-time

Mohammed VI Polytechnic University is an institution dedicated to research and innovation in Africa and aims to position itself among world-renowned universities in its fields

The University is engaged in economic and human development and puts research and innovation at the forefront of African development. A mechanism that enables it to consolidate Morocco’s frontline position in these fields, in a unique partnership-based approach and boosting skills training relevant for the future of Africa.

Located in the municipality of Benguerir, in the very heart of the Green City, Mohammed VI Polytechnic University aspires to leave its mark nationally, continentally, and globally.

1 Research Motivation

The problem of Simultaneous Localization and Mapping (SLAM) is one of the fundamental challenges in mobile robotics. Accurate environmental models are crucial for applications such as warehouse automation, digital twin creation, search and rescue operations, and agriculture. For instance, in warehouse automation, autonomous mobile robots use SLAM to navigate, retrieve items, and optimize storage, thereby improving efficiency and reducing human labor1 . Similarly, SLAM enables the creation of digital twins for virtual monitoring, helps rescue robots navigate hazardous environments, and supports precision agriculture by guiding autonomous farming equipment.

2 Problem Statement

Visual SLAM is a technique used in robotics and computer vision that enables a device, typically equipped with one or more cameras, to simultaneously build a map of an unknown environment and localize itself within that map using visual information. Unlike traditional SLAM methods that may rely on various types of sensors such as LIDAR, IMUs, or ultrasonic sensors, Visual SLAM primarily uses camera images to extract information about the surroundings. Despite significant progress in Visual SLAM, current solutions often face challenges related to error accumulation during camera tracking [4], handling occlusions and dynamic objects [1], scalability to large environments [3], and robustness to sensor noise (e.g., camera and depth sensors). These challenges can significantly impact the performance and reliability of SLAM systems, limiting their practical deployment in real-world scenarios.

3 Research Scope

Recent approaches addressing Visual SLAM use implicit neural scene representations such as Neural Radiance Fields (NeRF) [6] or explicit representations such as 3D Gaussian Splatting [2, 5]. These approaches yield impressive results, especially for 3D reconstruction. This research will initially explore these advanced techniques. Subsequently, the prospective candidate will investigate innovative solutions to improve the performance and robustness of SLAM systems regarding the aforementioned challenges. The proposed solution will be evaluated and compared to state-ofthe-art solutions on several indoor and outdoor datasets such as Scannet++2 and KITTI3 .

4 Admission Criteria

The PhD position is offered by the International Center of Artificial Intelligence of Morocco (Ai movement) at Mohammed VI Polytechnic University. Applicants with excellent academic records must hold a Master’s degree, an engineering degree, or an equivalent recognized degree in computer science or applied mathematics. Experience with 3D computer vision and robotics is desirable but not required. Additionally, candidates should possess excellent programming skills (Python and C++) and have good communication skills in English.

References [1] Tobias Fischer, Lorenzo Porzi, Samuel Rota Bulo, Marc Pollefeys, and Peter Kontschieder. ` Multi-level neural scene graphs for dynamic urban environments. arXiv preprint arXiv:2404.00168, 2024. [2] Bernhard Kerbl, Georgios Kopanas, Thomas Leimkuhler, and George Drettakis. 3d gaussian splatting for real-time radiance field rendering. ACM Transactions on Graphics, 42(4), July 2023. [3] Bernhard Kerbl, Andreas Meuleman, Georgios Kopanas, Michael Wimmer, Alexandre Lanvin, and George Drettakis. A hierarchical 3d gaussian representation for real-time rendering of very large datasets. ACM Transactions on Graphics, 43(4), July 2024. [4] Lorenzo Liso, Erik Sandstrom, Vladimir Yugay, Luc Van Gool, and Martin R Oswald. Loopy- slam: Dense neural slam with loop closures. arXiv preprint arXiv:2402.09944, 2024. [5] Hidenobu Matsuki, Riku Murai, Paul H. J. Kelly, and Andrew J. Davison. Gaussian Splatting SLAM. 2024. [6] Ben Mildenhall, Pratul P Srinivasan, Matthew Tancik, Jonathan T Barron, Ravi Ramamoorthi, and Ren Ng. Nerf: Representing scenes as neural radiance fields for view synthesis. Communications of the ACM, 65(1):99-106, 2021.

UM6P.