Research keywords
State estimation theory, Inertial/magnetic navigation, Kalman filtering, Cyber-physical security, Active defense, Artificial intelligence in navigation and mobility, Bearings-only target motion analysis and estimation, Navigation under cosmic radiation, Traffic state estimation, Large scale multimodal mobility networks, Shape estimation of continuum robots, Control of physical systems.
Summary of research work
Key challenges in navigation and their solutions
My research centers on advancing the field of navigation and estimation, with a particular focus on addressing the fundamental challenges that arise in real-world applications. I investigate attitude estimation, exploring both theoretical frameworks and practical solutions that improve the accuracy and reliability of rigid-body orientation measurements. This includes estimation using low-cost magnetic, angular rate, and gravity (MARG) sensors, either from standalone IMUs or integrated within smartphones, and energy-aware adaptive attitude estimation approaches that optimize performance under constrained resources. I also explore learning-based noise covariance matrix adaptation in Kalman filters for inertial navigation, enabling dynamic tuning of filter parameters to enhance robustness in uncertain and noisy environments. In addition, my research investigates learning-based mitigation of soft error effects on Kalman filter processing, as well as soft error assessment of attitude estimation algorithms running on resource-constrained devices under neutron radiation, ensuring resilience in harsh operational conditions. Beyond attitude, my research addresses velocity and position estimation in dynamic environments, including inertial velocity estimation for indoor navigation through magnetic gradient-based Kalman filters and learning models, pedestrian navigation using foot-mounted MARG sensors, and dead-reckoning techniques for marine animals, applying inertial navigation concepts to biological movement tracking. I also develop approaches for position, velocity, and attitude estimation based on MARG and position measurements under unknown inputs and outputs, maintaining navigation accuracy in partially observable, uncertain, and dynamically disturbed systems.
Modeling, estimation, and classification in urban and multimodal mobility systems
Recent research in urban transportation systems and traffic modeling emphasizes the integration of sensor-based monitoring, estimation and prediction techniques, machine learning, and data-driven estimation frameworks to enhance mobility, efficiency, and sustainability. In the domain of transportation mode detection (TMD), combining inertial (3-axis accelerometer and gyroscope) and barometric sensors enables robust classification of multiple travel modes across diverse sensor placements. A filtering is often applied to preprocess inertial signals, while feature extraction and selection tailored to sensor placement improve classifier robustness for previously unseen users. Machine learning models, including support vector machines (SVM), random forests (RF), gradient boosting (GB), and deep learning networks, are employed for sequence modeling and mode prediction, highlighting the importance of temporal dynamics in inertial data. My research also focuses on smartphone-based TMD applications that employ these models using inertial and magnetic data, while omitting global positioning system (GPS) to reduce energy consumption and preserve privacy. End-to-end pipelines involve data preprocessing, feature normalization, classification, and real-time client-server communication, ensuring reliable deployment in urban environments. Building on these sensor-based and data-driven approaches at the individual level, my research extends to city-scale systems, where modeling and estimation are applied to large-scale multimodal mobility networks to optimize urban transportation. I am working in that area, focusing on methods to analyze, predict, and optimize urban mobility across multiple transportation modes. My first results concern public transit modeling and incentives for large gatherings, where a traveller mobility model simulates mode choice based on routing apps, transit fare, and service frequency, while accounting for congestion, transit and road capacity, wait times, and parking availability. Using this framework, optimization techniques are applied to identify incentive strategies that balance operational cost and user satisfaction, promoting sustainable public transit use and improving urban mobility under high-demand scenarios.
Cyber-physical security in navigation and active defense strategies
Recent developments in cyber-physical system (CPS) security for navigation and control applications highlight the necessity of integrating robust state estimation, resilient control design, and active defense mechanisms to maintain system integrity under cyber threats. In attitude estimation using MARG sensors, a secure estimation framework formulated on the special orthogonal group SO(3) has been proposed to mitigate the impact of false data injection (FDI) attacks on sensor measurements. This approach leverages an invariant extended Kalman filter (IEKF) with an optimized Kalman gain matrix designed to minimize the upper bound of the state estimation error covariance, ensuring accurate attitude reconstruction even in adversarial environments. Within vehicular control systems, analyses of zero-dynamics attacks have demonstrated that malicious inputs exploiting the system’s internal model and unobservable subspaces can manipulate yaw rate and lateral acceleration without altering measurable outputs. Such stealthy perturbations emphasize the importance of observer-based detection schemes, redundant sensing architectures, and secure feedback design to preserve dynamic stability. Beyond passive protection, active defense strategies such as the Misleading Unauthorized Observer (MUO) technique introduce deceptive input–output signal modulation to disrupt unauthorized state estimation. By solving an optimization problem with undetectability constraints, the defender injects auxiliary control signals that increase the estimation error of eavesdropping observers while maintaining nominal system performance. Collectively, these advancements merge robust filtering, secure control theory, and adaptive deception techniques into a cohesive framework that enhances the resilience of modern navigation and vehicular CPS against sophisticated cyber-physical attacks.
Control problems in physical systems
My research focuses on advances in control systems that integrate predictive modeling, adaptive control, intelligent decision-making, and robust output-feedback synthesis to achieve high performance under uncertainty, dynamic disturbances, and nonlinearities. In industrial processes, such as phosphorite ore sintering, I investigate grey system-based predictive modeling for estimating the burn through point (BTP), combined with particle swarm optimization to synthesize real-time control laws. In aerospace, my work on adaptive control for spacecraft motion trajectories employs linearized nonlinear dynamics and the second Lyapunov method, with cascaded adaptation and feedback loops ensuring high-fidelity tracking and convergence of steady-state errors. In mobile robotics, I develop minimal rule-based fuzzy logic controllers for path planning and obstacle avoidance, combining goal-directed navigation with proactive collision avoidance. For switched linear and large-scale decentralized systems, my research designs robust H-infinity output-feedback controllers using descriptor redundancy, multiple Lyapunov functions, and linear matrix inequalities (LMIs) to ensure stability, disturbance attenuation, and tracking performance. Overall, my work demonstrates the integration of predictive, adaptive, and robust control to optimize performance, reliability, and resilience in industrial, aerospace, and robotic systems under complex and dynamic conditions.
Research projects
I have participated in numerous scientific projects at the local, national, and international levels, taking on responsibilities such as supervising postdoctoral researchers, PhD students, master's research students, and final-year engineering students' projects, as well as preparing project deliverables.
Improved bearings-only target motion analysis using AI tools (IMAnAI).
Type: International project.
Funding: CEFIPRA Indo-French call + INRIA défis program + Naval Group funding.
Period: 2023-2027.
Merging perception and quantitative measurements to assess quality of service in public transportation using machine learning (PercepTrans).
Type: International project.
Funding: Campus France, Partenariat Hubert Curien Franco Sud-Africain (PHC PROTEA)
Period: 2021-2023.
Scale-freeback (SFB).
Type: International project.
Funding: European Research Council (ERC) H2020-EU.1.1.
Period: 2016-2022.
Scalable proactive event-driven decision-making (SPEEDD).
Type: International project.
Funding: FP7-ICT.
Period: 2014-2017.
Continuous methods for the control of large networks (COCOON).
Type: National project.
Funding: ANR AAPG 2022.
Period: 2022-2027.
Tachymètre magnéto-inertiel couplé vision (TMI-V).
Type: National project.
Funding: Direction Générale de l’Armement (DGA), RAPID program.
Period: 2018-2022.
Fusion autonome de capteurs optroniques-électromagnétiques-inertiels pour la navigation du système de combat aérien future (FAUCON) - Partie A : navigation et géolocalisation par vision.
Type: Industrial project.
Funding: Safran électronique & défense - Thales avionics, Direction générale de l’armement (DGA).
Period: 2021-2022.
Enhancing surgery with deep learning-controlled continuum robots (INSPECT).
Type: Local project.
Funding: Multidisciplinary Institute in Artificial Intelligence Grenoble Alpes (Chaire MIAI CLUSTER)
Period: 2025-2029.
Robots for real word interaction (BOOT).
Type: Local project.
Funding: IDEX University Grenoble Alpes.
Period: 2022-2025.
Détection fiable des modes de déplacement pour la navigation hybride en mobilité urbaine douce (MOBIDOU).
Type: Local project.
Funding: Multidisciplinary Institute in Artificial Intelligence Grenoble Alpes (MIAI).
Period: 2023-2025.
AI and dynamical systems: new paradigms for control and robots (AIBot).
Type: Local project.
Funding: Multidisciplinary Institute in Artificial Intelligence Grenoble Alpes (Chaire MIAI).
Period: 2019-2023.
Capture et analyse d’activités humaines par modules inertiels : vers une solution adaptée à la navigation multimodale urbaine intelligente (CAPTIMOVE).
Type: Local project.
Funding: IDEX University Grenoble Alpes.
Period: 2019-2021.
Plateforme expérimentale de modules inertiels pour la capture et l’analyse fine du mouvement humain pour une reconnaissance fiable des modes de transport utilisés (PLATEFORME).
Type: Local project.
Funding: IDEX University Grenoble Alpes.
Period: 2020.
Nanosatellite project: Advanced modelling and control of attitude dynamics for quantum communication (SPACE).
Type: Local project.
Funding: IDEX University Grenoble Alpes.
Period: 2018-2019.
Capture de mouvements humains par centrales inertielles/d’attitude et smartphones : Vers l’analyse d’anomalies neurologiques et fonctions motrices (POSTURE).
Type: Local project.
Funding: IDEX University Grenoble Alpes.
Period: 2018.
Localization techniques for pedestrian navigation based on inertial measurements in indoor environments (LOCATE-ME).
Type: Local project.
Funding: LabEx Persyval-Lab University Grenoble Alpes.
Period: 2014-2015.
Contrôle-commande des éoliennes offshore flottantes.
Type: Industrial project.
Funding: BW Ideol company.
Period: 2012.
Platforms (P), Android applications (A), Database (DB), and Benchmark (B)
[P1] GTL-V: Experimental platform for Grenoble Traffic Lab-Ville research. This platform was developed and used as part of my PhD student M. Rodriguez Vega’s doctoral thesis (2017-2021).
[A1] Senslogs: Android application for recording and storing data from all smartphone sensors (Accelerometer, gyroscope, GPS, barometer, magnetometer, etc.). This application was developed and used as part of my PhD student T. Michel’s doctoral thesis (2014-2017).
[A2] TyrAr, AmiAr, Login'AR: Geo-localized augmented-reality Android browser for three applications in indoor and outdoor navigation: TyrAr for improved attitude estimation, AmiAr for augmented reality in an apartment, and Login’AR for augmented reality in the INRIA Grenoble showroom. This browser was developed during my PhD student T. Michel’s doctoral thesis (2014-2017).
[A3] MOBIDOU: Android application that collects and consolidates readings from the smartphone’s various sensors (Accelerometer, gyroscope, GPS, and magnetometer) into a single file, recording the date and time of each acquisition, has been developed. The processed data is then analyzed by an offline classification algorithm that calculates the proportions of urban travel modes used (Public transport, bicycle, kick-scooter, walk) over the time period selected by the user. This application was developed and used by my postdoctoral researcher, I. Gharbi (2023–2025). A second version of this application is currently in development.
[DB1] PercepTrans Data: A database comprising raw sensor data collected from Inertial Measurement Units (IMUs) of Smartphones during 113 transportation trips in Johannesburg and Durban, South Africa. The dataset characterizes vehicle movement across multiple transportation modes, including cars, minibus taxis, buses, and ride-hailing services. Passengers carried the sensors with minimal movement, ensuring the recorded signals reflect the vehicle’s motion. The data supports analyses of vehicle dynamics, driving quality, and discrepancies between perceived and actual driving performance.
[DB2] TMD-CAPTIMOVE: A database bringing together inertial and pressure measurements collected from inertial units (accelerometer, gyroscope, GPS, pressure, Earth’s magnetic field, etc.) attached to the human body, recorded from 34 volunteers during a multimodal travel scenario (Public transport, bicycle, kick-scooter, walk). This database was created by my postdoctoral researcher, F. Taia Alaoui (2019-2021).
[B1] On Attitude Estimation with Smartphones (AOES): A dedicated benchmark for estimating smartphone attitude (3D orientation) using observers and Kalman filters, intended for geo-localized augmented reality applications. This benchmark was developed and used as part of my PhD student T. Michel’s doctoral thesis (2014-2017).