Deep Learning-Based Gait Mode Classification to Enhance Adaptive Control of Robotic Assistance

Université Paris Est Créteil; Les Ulis; Sur place; EUR 40 000 - 60 000; Plein temps

Résumé du poste

Ce projet de doctorat au Laboratoire LISSI cherche à développer une assistance robotique intelligente pour aider les personnes âgées ou celles ayant des difficultés de marche. L'enjeu est d'adapter en temps réel l'assistance fournie grâce à des systèmes de contrôle avancés, visants à améliorer la sécurité et la mobilité des utilisateurs. Une expertise en robotique et en contrôle adaptatif est essentielle pour mener à bien ce projet novateur financé par des fonds publics.

Qualifications

• Expérience en robotique et contrôle dynamique souhaitée.
• Compétences en traitement de données et en apprentissage automatique.
• Capacité à travailler avec des systèmes de contrôle complexes.

Responsabilités

• Développer un système portable d'assistance robotique basé sur l'apprentissage profond.
• Gérer les incertitudes et les variations des conditions réelles.
• Intégrer les architectures de contrôle robustes et flexibles.

Connaissances

Apprentissage profond

Contrôle adaptatif

Robotique wearable

Analyse de données

Formation

Doctorat en robotique, intelligente ou domaine connexe

Description du poste

This PhD project, conducted at the Laboratoire Images, Signaux et Systèmes Intelligents (LISSI, EA ), addresses the challenge of providing intelligent robotic assistance to frail individuals with reduced locomotor capacity, such as older adults or people affected by gait impairments like foot-drop. These individuals often experience significant difficulties in walking safely and efficiently across diverse terrains—such as stairs, ramps, or uneven surfaces—which heightens the risk of falls and severely limits autonomy. The central goal of this project is to develop a wearable system combining deep learning-based gait mode classification with advanced control strategies to deliver context-aware, real-time robotic assistance via an active orthosis. The proposed system aims to enhance safety, reduce physical strain, and improve mobility performance in daily life scenarios.

Sensor data will be acquired through inertial measurement units (IMUs) and force-sensitive resistors (FSRs) placed unobtrusively on or in the user’s footwear. Deep learning models will be trained to recognize gait modes in real time, providing the necessary contextual information to guide assistive torque generation. However, the core focus of the research lies in the development of an adaptive control scheme capable of translating this information into smooth, stable, and responsive actuation, even under uncertain and variable conditions. The controller must account for sensor noise, time delays, biomechanical variability among users, and the unpredictable nature of real-world environments—including the presence of obstacles and terrain-induced perturbations.

Building upon the SIRIUS team’s expertise in human-in-the-loop control, adaptive and sliding-mode controllers, and impedance shaping for wearable robots, the project will explore control architectures that integrate robustness and flexibility. A particular emphasis will be placed on managing system uncertainties and external disturbances, ensuring that the robotic orthosis can modulate its assistance appropriately without destabilizing the user or disrupting natural movement. The expected outcome is a real-time, embedded solution capable of delivering assistive torques that are both physiologically coherent and dynamically adapted to the walking context—contributing to the development of next-generation, intelligent rehabilitation technologies for frail populations.

Funding category

Public funding alone (i.e. government, region, European, international organization research grant)