Job Offer

SUMMARY (PhD Offer)

Digital systems are the main focus of research on mobility aids for visually impaired people: GPS guidance applications on smartphones, instrumented blind canes, auditory description of the environment, etc., and to a lesser extent on the assistance provided by guide dogs.  The latter research focuses on animal mediation [1], training adult or minor beneficiaries [2], and selecting and training dogs [3-4]. The matching of guide dogs and beneficiaries, which plays a crucial role in the success of mobility assistance, has received little research attention. The current process relies mainly on qualitative data, such as assessing beneficiary needs, dog profiling and supervised meetings. This approach can lead to incompatibilities that affect the effectiveness and well-being of the duo. In this context, this thesis aims to evolve this matching process by introducing qualitative and automatable approaches with a tool for accurately measuring dog-human compatibility based on behavioural, emotional and biomechanical criteria as well as walking patterns learned from dogs and recipients.

This thesis is part of an international collaboration between MIRA in Canada, a key player in training guide and assistance dogs, the Valentin Haüy Association, a long-standing player in helping the visually impaired in France, and the Gustave Eiffel University. The thesis will develop a methodology to collect precise data on the walking styles of the beneficiaries and the dogs, as well as their behavioural and emotional states, using wearable motion sensors and advanced behavioural analysis tools from INMOB labcom [5]. The research will start by developing and adapting motion capture and behavioural analysis devices for easy and effective use with dogs and beneficiaries, and by developing and integrating acquisition systems. During pairing sessions and in a variety of situations, gait, behavioural and emotional data will be collected to create detailed profiles. The collected data will be processed and analysed to develop a dog-beneficiary matching model based on compatibility of gait and behavioural profiles. Finally, the model will be integrated into the MIRA matching process for evaluation and adjustment based on feedback and matching success. 

This work will mobilise interdisciplinary skills in geomatics, etiology, and artificial intelligence to process the collected data and integrate it into a decision-support tool. The methodology will involve capturing movement using inertial sensors, analyzing gait dynamics using signal processing algorithms, and observing and quantifying behaviour during human-dog interactions.

The expected benefits of this project are manifold. The research will have an impact on the quality of dog-recipient matches, with a reduction in the number of failures and re-assignments. It will also contribute to an improvement in the well-being of both the dogs and the recipients by ensuring a better match right from the selection stage. Finally, scientific advances are expected in understanding the dynamic interactions between assistance dogs and humans, particularly regarding gait and behaviour. This research will also strengthen interdisciplinary collaboration between the fields of ethology and movement analysis, and illustrate the potential of integrated approaches to solving complex problems. The work will be carried out in a framework of academic, industrial and associative collaboration, with a strong potential for technology transfer to organisations involved in the training of guide and assistance dogs.

REFERENCE

[1] Alice Mignot, « Bénéfices de l'approche pluridisciplinaire dans la compréhension de la médiation animale ainsi que dans l'évaluation du bien-être du chien médiateur », thèse doctorale, Université de Nanterre - Paris X, 2022. 

[2] Fanny Menuge, Identification of factors influencing the acquisition of skills necessary to work as a guide dog for blind children : conception d’un programme d’optimisation. Animal biology. Institut National Polytechnique de Toulouse - INPT, 2022. English.

[3] Fondation MIRA,  Chiens-guides pour les personnes non-voyantes ou ayant une déficience visuelle https://www.mira.ca/fr/programmes/chien-guide-deficience-visuelle

[4] Nicolas Dollion; Margot Poirier; Association Handi'Chiens; Fondation Mira; Florian Auffret; Nathe François; Pierrich Plusquellec; Marine Grandgeorge. (2024) "Effects of service dogs on children with ASD’s symptoms and parents' wellbeing: on the importance of considering those effects with a more systemic perspective.". PLoS One 2024 Jan 3;19(1) https://doi.org/10.1371/journal.pone.0295702     

[5] Publications du labcom INMOB (cartographie du handicap par mesure INertielle pour faciliter la MOBilité) financé par l’ANR, https://anr.hal.science/search/index/?q=*&anrProjectReference_s=ANR-20-LCV1-0002 

KEYWORDS

Navigation, Mobility assistance, Guide dogs, Sustainable mobility

The complexity of positioning scenarios and the requirements for location-based applications have significantly increased due to technological advancements, urbanization, and the broadening of use cases across industries. Applications in fields such as autonomous vehicles, indoor navigation, healthcare, augmented reality, and logistics all require increasingly sophisticated solutions to deliver precise, reliable, and adaptive experiences that account for dynamic and diverse conditions. At the same time, vast amounts of navigation data are generated from different navigation sensors, necessitating advanced methods for processing, analyzing.   

That is why, data-driven approaches are becoming increasingly popular in the localization and navigation domain due to their ability to address complex challenges that traditional methods struggle to model accurately. For instance, GNSS multipath and Non-Line-of-Sight (NLOS) reception errors are difficult to predict with conventional models, leading many researchers to explore Artificial Intelligence (AI)-based techniques for detecting and excluding faulty GNSS measurements [1-2]. AI models, particularly those leveraging machine learning, offer a more adaptive and dynamic approach to identifying and mitigating these errors, significantly enhancing the reliability of GNSS-based positioning systems. Additionally, parameters for navigation filters, such as the Extended Kalman Filter (EKF) and Factor Graph Optimization (FGO), are crucial for determining positioning performance, as they are highly sensitive to contextual factors like environmental conditions and device quality. Fine-tuning these filtering parameters can greatly improve the stability and robustness of location accuracy, and deep learning techniques show promise in automating and optimizing this process, making it more efficient and precise [3]. In the navigation and positioning domain, supervised learning techniques are often applied to ensure more controllable, predictable results, and better overall performance, providing an effective means to adapt to diverse conditions and system requirements. 

The labels required for supervised learning, particularly high-quality labels, are often costly to obtain due to factors such as the need for human labor, expensive equipment, or access to detailed map data. For example, labeling GNSS Line-of-Sight (LOS) and NLOS data typically relies on cumbersome devices like fish-eye cameras or 3D map-aided systems, which can introduce errors such as image segmentation inaccuracies, camera calibration errors, and map discrepancies. The main research question and objective of this PhD thesis is to propose a generalizable data labeling methodology that uses reinforcement learning (RL) to automatically generate high-quality labels. By utilizing only the reference trajectory, this approach aims to label intermediate quantities in particular sensor data quality and filtering parameters in a cost-effective and scalable way. 

Reinforcement learning is particularly suited for this task because it enables an adaptive, autonomous learning process where the system can continuously refine its labeling strategy based on feedback from the environment. The main tasks in this PhD thesis include: 1) A thorough state-of-the-art on reinforcement learning focusing on its applications in localization and navigation systems, as well as its potential to address challenges in data labeling. 2) Modeling the navigation data labeling problem within the framework of reinforcement learning by carefully defining the core elements of the problem, including the state, actions, environments, and rewards. This leads to an appropriate methodology that allows reinforcement learning algorithms to autonomously optimize the labeling process, relying only on the reference trajectory, which is anyway needed for performance evaluation. In this way, the cost and efficiency of data labeling will be significantly reduced.  3) Two particular study will be conducted to demonstrate the effectiveness of the proposed methodology: a) sensor quality labeling, eg., GNSS and Inertial Navigation System (INS) data; b) filtering parameter labeling, e.g., EKF and/or FGO, ensuring that the filter remains robust and accurate under varying conditions.

Required capacity: reinforcement learning, signal processing, applied mathematics, multisensory fusion positioning, state estimation. 

Keywords

Positioning, multisensory data, labeling, reinforcement learning 

[1] N. Zhu, R. He, Z. Wang, CarNet: A generative convolutional neural network-based line-of-sight/non-line-of-sight classifier for global navigation satellite systems by transforming multivariate time-series data into images, Engineering Applications of Artificial Intelligence, Volume 145, 2025, 110160, ISSN 0952-1976, https://doi.org/10.1016/j.engappai.2025.110160.

[2] García Crespillo, O., Ruiz-Sicilia, J. C., Kliman, A., & Marais, J. (2023). Robust design of a machine learning-based GNSS NLOS detector with multi-frequency features. Frontiers in Robotics and AI, 10, 1171255.

[3] Li, Shuo, et al. "Exploring the Potential of Deep Learning Aided Kalman Filter for GNSS/INS Integration: A Study on 2D Simulation Datasets." IEEE Transactions on Aerospace and Electronic Systems (2023).

[4] Zhu, N., Bouronopoulos, A., Leduc, T., Servières, M., & Renaudin, V. (2023, April). Evaluation of the Human Body Mask Effects on GNSS Wearable Devices for Outdoor Pedestrian Navigation Using Fisheye Sky Views. In 2023 IEEE/ION Position, Location and Navigation Symposium (PLANS) (pp. 841-850). IEEE.

Possibility to continue as a PhD student at the GEOLOC lab

Subject:

In urban environments, the accuracy of Global Navigation Satellite Systems (GNSS) is severely compromised by the environment. Buildings, vegetation and city morphology can cause reflections, diffractions or blockages of satellite signals. This degradation of measurements (pseudorange, carrier phase, doppler shift) has a significant impact on positioning quality - a critical issue for high-precision applications such as autonomous vehicles and guidance for the visually impaired.

To estimate a position from satellite measurements, snapshot algorithms such as Weighted Least Squares (WLS) regression are commonly used. One strategy for improving accuracy is to assign a confidence indicator to each satellite for each instant of measurement. These indicators are then used as weights in the position estimation, reducing the influence of lower-quality measurements. Although heuristics exist in the literature, they require hyperparameters tuning and are not necessarily suited to complex urban environments.

This internship proposes to explore a new approach based on Reinforcement Learning. The aim is to determine a function capable of predicting a satellite's confidence indicator from its measured signals. Reinforcement Learning is particularly well suited to this problem, as it does not require a ground truth for the weights. Instead, we only have the sensor's reference position. In this paradigm, an agent predicts weights and receives a reward based on the quality of its predictions, here evaluated with the positioning error made with the predicted weights.

Tasks:

- Literature review on Reinforcement Learning techniques

- Modeling of the learning problem

- Implementation and training of a model, using data collected by the GEOLOC lab

- Evaluation of the model and comparison with an end-to-end deep learning paradigm (https://arxiv.org/abs/2409.12996)

Skills required:

• Experience in Machine Learning
• Coding (Python or MATLAB)

Context:

The GEOLOC laboratory is located on the Nantes campus of the Gustave Eiffel University. The laboratory's research focuses on geolocation with multi-sensor fusion for multimodal transport.

Details:

• The internship will take place at the GEOLOC laboratory, on the Nantes campus of the Gustave Eiffel University (Allée des Ponts et Chaussées, 44340 Bouguenais).
• 600€ - 700€ / month, depending on the number of days of attendance.
• Duration: 5 to 6 months
• Desired start date: between February and April 2025