From Pixels to Commands: End-to-End Learning for Robot Navigation

We are looking for a highly motivated PhD candidate to join our research team to work on the exciting topic of agile and safe navigation for mobile robots using vision and inertial sensors. The successful applicant will be part of the Intelligent Systems research area at the Institute for Research in Technology (IIT) and will collaborate closely with the Department of Electronics, Control and Communications (DEAC) at Comillas Pontifical University.

The core objective of the research is to develop end-to-end learning frameworks that enable mobile robots -both ground and aerial- to navigate autonomously and reliably in unstructured and dynamic environments, such as forests, disaster zones, or industrial sites. Unlike traditional robotics pipelines that rely on modular architectures for perception, mapping, and control, the focus of this PhD will be to jointly learn these components using deep learning techniques, directly mapping visual and inertial data to low-level control commands.

Current methods for robot navigation often require heavy pre-processing, hand-crafted features and algorithms, or maps that are difficult to obtain or maintain in complex, real-world environments. This thesis aims to overcome these limitations by training robust, task-driven models capable of generalizing across different settings and operating conditions.

The candidate will explore topics such as deep reinforcement learning, visual-inertial sensor fusion, and domain adaptation, with an emphasis on practical deployment and safety following a sim-to-real approach.
If you are passionate about robotics, AI, and building real-world systems that learn and adapt, we encourage you to apply.