D5 - Signal and digital images, robotics

Person in charge
Éric MARCHAND (Professor Université de Rennes 1)
Description

Digital Signal & Images, Robotics

The Digital Signals & Images, Robotics (D5) Department contributes research on the formalization, design and experimentation of models and algorithms to process, transmit and transform digital signals and data stemming from a variety of physical phenomena, with a focus on acoustic and audio signals, video and visual data, motion signals and captures, medical images, geoscience observations, and physical embodiment and interaction with the environment (robotics).

The main results concerns the following broad topics:

  • Advanced sensor-based control in robotics and shared control architectures: trajectory planning and maneuvering for UAVs; decentralized algorithms for multi-robot formation control; new wearable haptic interfaces as well as haptic rendering techniques; shared control algorithms for telemanipulation with a dual-arm system; semi-autonomous navigation of a robotic wheelchair for impaired people
  • Machine learning, sparse modeling and compressive sensing: establishing the theoretical and algorithmic foundations of compressive statistical learning; compressive sensing framework for reconstructing a light field from a single-sensor; novel sampling schemes based on dictionary learning and compressed sensing; optimal transport metrics in graph signal processing, domain adaptation in deep neural networks; out-of-sample generalization manifold learning and clustering on manifolds
  • Optimal and uncertainty-aware sensing: model-based visual tracking methods; active sensing to analyze and synthesize optimal trajectories for a robotic system; new control laws for visual servoing based on deep learning and model-free deformation servoing method for soft bodies
  • Audio, signal and image processing: novel approaches for source localization and separation; new models and learning paradigms to cope with complex data and tasks in earth observation;  light field compression and reconstruction; development of a unique experimental and methodological framework for neurofeedback ; applications in neuroradiology and neurological disorders