Advancing SLAM through Deep Learning

Niclas Vödisch (Ph.D. Student)

A fundamental challenge in the field of mobile robotics is to determine the robot’s pose relative to its environment. Commonly, this is achieved by employing simultaneous localization and mapping (SLAM) systems. While classical methods are well studied and provide reliable solutions to the localization problem, how to exploit the advances made in machine learning is still an ongoing question. In my Ph.D. project, I investigate how adding components based on deep learning to the concept of SLAM can improve both robustness and efficiency. For instance, extending geometry-based methods by semantic and learning-based features can aid odometry computation and loop closure detection. Additionally, learnable representations of map data can be leveraged to deal with temporary changes in the environment, such as dynamic objects or road work, as well as with more long-term alterations, e.g., due to seasonal conditions.

Primary Host:	Wolfram Burgard (Technical University of Nuremberg)
Exchange Host:	Davide Scaramuzza (ETH & University of Zürich)
PhD Duration:	01 June 2021 - 31 May 2025
Exchange Duration:	01 June 2023 - 30 November 2023 - Ongoing