Chercheur Doctorant Semantic Data Compression for Extended Reality in Aeronautic Industry H/F

Le profil souhaité

Scientific Fields

•    Extended Reality Systems & Human–Computer Interaction (HCI)

•    Edge–Cloud Computing and Distributed Systems

•    Industrial Internet of Things (IIoT)

•    Signal Processing and Semantic Data Compression

Keywords

•    Data Compression

•    Extended Reality (XR)

•    Edge–Cloud Computing

•    Real-Time Streaming

•    Machine Learning

SupervisionThesis Director

•    Samir OUCHANI, PhD., HDR., Research Director, Aix-en-Provence, France.

Thesis Supervisor

•    Hugues Marie KAMDJOU, PhD., Associate Professor

Acknowledgment

This work is conducted as part of Campus Aero Adour (C2A) project funded by the government under the France 2030 Plan.

1

Research WorkAbstract

Aeronautical eXtended Reality (XR) deployments (maintenance assistance, inspection, and training) increasingly rely on rich 3D scenes, CAD/digital-twin assets, and multi-sensor streams that must be delivered in real time to mobile headsets. In operational settings, network connectivity is often unstable (hangars, tarmacs, factories), while XR devices remain constrained in compute, memory, battery, and network bandwidth. Classical compression and streaming reduce bitrate but treat all content uniformly, failing to prioritize task-critical elements (e.g., safety-relevant parts, occluded components, procedural cues). This gap motivates semantic, edge-assisted data compression strategies that reduce communication cost while preserving the information that directly impacts industrial performance and user effectiveness.

This thesis addresses trustworthy semantic compression to enable real-time XR in aeronautics under strict bandwidth, latency, and device-energy constraints. The method proposes task-aware representations that prioritize the transmission of operational meaning over uniform raw pixels or geometry. In this approach, what matters encompasses critical objects, procedural cues, real-time IIoT data essential for situated guidance, but also discrepancies between the real system and its digital twin, specifically highlighting structural discrepancies such as component presence/absence or geometric deviations. Compression decisions are optimized and executed at the edge to reduce end-to-end delay and stabilize user-perceived quality on resource-limited XR headsets [1, 2]. The resulting pipeline couples semantic inference with adaptive coding and streaming to minimize transmitted data while preserving operational accuracy.

PhD projectScientific Context

Industry 5.0 promotes human-centric, resilient, and sustainable production systems, where XR supports situated guidance, remote collaboration, and digital-twin interaction in safety-critical domains such as aeronautics. Achieving stable and low-latency XR requires continuous pose tracking, 3D scene updates, and multi-sensor fusion, often under fluctuating network conditions and strict operational constraints. Yet, current head-mounted displays (HMDs) remain limited in CPU/GPU capability, memory, energy, and thermal budget, which restricts on-device inference and high-fidelity rendering at scale. Consequently, conventional end-to-end pipelines that stream dense geometry, textures, and video frames are prone to bandwidth bottlenecks and motion-to-photon delays that degrade user performance and comfort in dynamic Industrial IoT (IIoT) environments [3, 4].

Semantic data compression has emerged as a promising direction to bridge this gap by encoding meaningful structures (objects, parts, relations, and task-relevant regions) rather than uniformly compressing raw signals. Recent studies show that semantic-aware representations can significantly improve transmission efficiency in industrial settings by prioritizing critical content and reducing redundancy [5, 6]. In this thesis context, deep learning-driven semantic analysis is leveraged to identify what is operationally important (e.g., aircraft components, procedural cues, hazard zones) and to drive adaptive compression and streaming decisions. Beyond bitrate reduction, this paradigm aims to preserve task accuracy while improving end-to-end responsiveness, thereby enhancing both Quality of Service (QoS) and Quality of Experience (QoE) for XR users in aeronautical workflows [2, 7].

Thesis subject

This PhD thesis targets the design of a robust semantic data compression pipeline for real-time XR in IIoT, with a particular focus on aeronautical workflows. The thesis will deliver an end-to-end method that (i) models and extracts semantics from XR scenes/datasets, (ii) performs adaptive compression/aggregation driven by these semantics, and (iii) validates the approach under realistic network and device constraints. The key objectives of this PhD thesis include:

•    Semantic workload characterization: Analyze the semantic data generated by aeronautic industrial XR

pipelines (poses, object states, part-level relations, interactions, annotations, and context) to determine

what must be preserved for task performance and what can be reduced through compression or aggregation.

•    Semantic-aware compression and aggregation: Design and implement efficient Machine Learning (ML) algorithms that jointly exploit semantic structure and temporal redundancy to optimize the uplink/downlink between HMDs and edge servers. This includes learning-based importance prediction (e.g., task-aware ROI/part importance), semantic quantization, and progressive/partial transmission strategies.

•    Edge-assisted decision making: Investigate edge/cloud orchestration where semantic inference and compression policy selection are offloaded to edge servers, reducing HMD compute/energy usage while meeting real-time constraints (motion-to-photon delay, frame stability, and reliable pose updates).

•    Rigorous evaluation and benchmarking:   Conduct extensive experiments to assess (i) communication efficiency (bitrate, packet loss sensitivity), (ii) QoS (end-to-end latency, frame rate stability, pose/estimation accuracy), (iii) server-side cost (CPU/GPU utilization, memory footprint, scalability), and (iv) user- centric QoE (perceived quality, comfort, task completion time/error rate) in IIoT-like conditions.

•    Prototype integration in an aeronautical XR setting: Integrate the proposed methods into an XR prototype (HMD + edge) and validate performance on representative aeronautical scenarios (maintenance assistance, inspection guidance, or training), demonstrating deployability and measurable gains over baseline codecs and non-semantic compression.

Previous Work in the Laboratory

In this context, several research works and projects have been conducted in CESI LINEACT laboratory. These include, among others:

•    Resource-constrained eXtended Reality operated with digital twin in IIoT [1], [8].

•    JENII project, is a remote learning initiative for the future industry, built upon immersive and collaborative environments centered around digital twins of real industrial systems.

These foundations will be crucial for integrating semantic data compression into XR applications, as the team has experience with both XR technologies and data processing frameworks.

Work Program

The PhD will be structured over three years with incremental deliverables, from foundations to prototype validation:

•    Year 1 — State of the art & problem formalization: Get familiar with our existing tools and experimenting XR environment, perform an in-depth literature review (XR streaming, semantic data compression, edge offloading, QoS/QoE), define the system model and evaluation protocol, and develop a first semantic- compression framework prototype (baseline + semantic extraction module). Deliverable: initial benchmark dataset/pipeline and one conference/journal submission.

•    Year 2 — Method design & implementation: Design and implement core algorithms (semantic importance prediction, semantic quantization/aggregation, progressive transmission, edge-assisted policy selection), and validate them in controlled/simulated XR settings under variable network conditions. Deliverables: open experimental framework and submitting the results to top venues.

•    Year 3 — Integration, large-scale evaluation & thesis writing: Integrate the approach into an end-to- end XR+edge prototype, conduct extensive performance and user/task evaluations (QoS, QoE, scalability, robustness), optimize deployment aspects, and finalize the dissertation. Deliverables: final framework release, one/two additional submissions, and thesis defense.

Expected Scientific/Technical Output

The expected outcomes of this thesis include:

•    Novel methods: design and validation of original semantic-aware compression and aggregation algorithms tailored to real-time XR streams in IIoT/aeronautical contexts (e.g., task-aware ROI/part importance, progressive semantic transmission, edge-assisted policy selection).

•    Prototype and reproducibility: an end-to-end XR–edge prototype demonstrating measurable gains over non-semantic baselines, accompanied by a reproducible evaluation pipeline (datasets/scenarios, benchmarks, and experimental scripts when possible).

•    Scientific publications: at least three peer-reviewed publications (Class A or Q1 journals and/or top-tier conference proceedings) in areas such as semantic compression, XR systems, edge/cloud computing, and IIoT.

•    Dissemination: presentation of the research at a minimum of two international conferences/workshops, including live demonstrations when feasible.

•    Technology transfer: technical reports and integration guidelines facilitating adoption in industrial XR workflows (maintenance/inspection/training), with documentation of deployment constraints and best practices.

Laboratory Presentation

CESI LINEACT (UR 7527), Laboratory for Digital Innovation for Businesses and Learning to Support the Competitiveness of Territories, anticipates and accompanies the technological mutations of sectors and services related to industry and construction. The historical proximity of CESI with companies is a determining element for our research activities. It has led us to focus our efforts on applied research close to companies and in partnership with them. A human-centered approach coupled with the use of technologies, as well as territorial networking and links with training, have enabled the construction of cross-cutting research; it puts humans, their needs and their uses, at the center of its issues and addresses the technological angle through these contributions. Its research is organized according to two interdisciplinary scientific teams and several application areas:

•    Team 1 « Learning and Innovating » mainly concerns Cognitive Sciences, Social Sciences and Management Sciences, Training Techniques and those of Innovation. The main scientific objectives are the understanding of the effects of the environment, and more particularly of situations instrumented by technical objects (platforms, prototyping workshops, immersive systems…) on learning, creativity and innovation processes.

•    Team 2 « Engineering and Digital Tools » mainly concerns Digital Sciences and Engineering. The main scientific objectives focus on modeling, simulation, optimization and data analysis of cyber physical systems. Research work also focuses on decision support tools and on the study of human-system interactions in particular through digital twins coupled with virtual or augmented environments.

These two teams develop and cross their research in application areas such as:

•    Industry 5.0,

•    Construction 4.0 and Sustainable City,

•    Digital Services.

Areas supported by research platforms, mainly that in Rouen dedicated to Factory 5.0 and those in Nanterre dedicated to Factory 5.0 and Construction 4.0.

Positioning in Laboratory Research Themes

The PhD topic is related to the research works conducted within the CESI LINEACT Engineering and Digital Tools team. The thesis aims to harness the synergies between the research areas of Resilient and Safe Systems (R2S) and digital tools within team 2 of CESI LINEACT.

Presentation of C2A project

Supported by state investment as part of the France 2030 Plan, « Campus Aero Adour » (C2A) is a project to support the digital and environmental transition of the aeronautics industry in the Adour territory. As a laureate of the « AMI Compétences et Métiers d’Avenir » call for projects under the ’Producing Low-Carbon Aircraft’ strand, C2A will benefit from State support through the France 2030 initiative over five years.

Thesis Organization

•    Funding: France 2030.

•    Workplace: CESI LINEACT, 6 Rue Saint-John Perse, 64000 Pau, France.

•    Start Date: October 2026.

•    Duration: 3 years.

Recruitment Terms and ConditionsModalities

Application review and interview. All eligible applicants are encouraged to apply. Applications will be processed as they arrive, early application is highly encouraged. Application should include:

•    Detailed Curriculum Vitae of the candidate. In case of a break in academic studies, please provide an explanation;

•    A motivation letter explaining your motivations for pursuing a PhD thesis;

•    Transcripts of Master 1 and Master 2 or current year of M2, and/or corresponding grade reports;

•    BSc/MSc/Ing. certificates (or equivalent level);

•    Two recommendation letters at least.

Skills

The candidate should possess a Master student or equivalent in Computer Science or Applied Mathematics. She/He should have some knowledge and experience in a number of the following points:

•    Scientific and Technical Skills:

–    Solid programming and software development tools skills (Docker, Python, Node JS, C# and Unity 3D),

–    Strong interest in XR technologies, digital twins, edge computing, and cloud architecture,

–    Familiarity with data compression, Machine Learning, networking concepts and protocols, particularly in the context of edge computing and cloud architecture,

–    Effective communication skills in English/French and the ability to collaborate within a multidisciplinary team environment.

•    Interpersonal Skills:

–    Being autonomous, having initiative and curiosity,

–    Ability to work in a team and have good interpersonal skills,

–    Being rigorous.

References

[1]    Hugues M. Kamdjou, David Baudry, Vincent Havard, and Samir Ouchani. Resource-constrained extended reality operated with digital twin in industrial internet of things. IEEE Open Journal of the Communications Society, 5:928–950, 2024.

[2]    Bowen Zhang, Zhijin Qin, and Geoffrey Ye Li. Semantic communications with variable-length coding for extended reality. IEEE Journal of Selected Topics in Signal Processing, 17(5):1038–1051, 2023.

[3]    Chaowei Wang, Yehao Li, Feifei Gao, Danhao Deng, Jisong Xu, Yuhan Liu, and Weidong Wang. Adaptive semantic-bit communication for extended reality interactions. IEEE Journal of Selected Topics in Signal Processing, 17(5):1080–1092, 2023.

[4]    Heodoros Theodoropoulos, Antonios Makris, and et al. Abderrahmane Boudi. Cloud-based xr services: A survey on relevant challenges and enabling technologies. Journal of Networking and Network Applications, 2:1–22, 2022.

[5]    Sifat Rezwan, Huanzhuo Wu, Juan A. Cabrera, Giang T. Nguyen, Martin Reisslein, and Frank H. P. Fitzek. cxr+ voxel-based semantic compression for networked immersion. IEEE Access, 11:52763–52777, 2023.

[6]    Xinyi Tu, Riku Ala-Laurinaho, Chao Yang, Juuso Autiosalo, and Kari Tammi. Architecture for data-centric and semantic- enhanced industrial metaverse: Bridging physical factories and virtual landscape. Journal of Manufacturing Systems, 74:965–979, 2024.

[7]    Hanh T.M. Tran, Hieu V. Nguyen, Van-Phuc Bui, Tien Ngoc Ha, Van Tho Nguyen, Duc-Hien Nguyen, and Mai T.P. Le. Encoding reality with semantic interpretation in metaverse interactions. AEU – International Journal of Electronics and Communications, 187:155512, 2024.

[8]    Hugues M. Kamdjou and Samir Ouchani. A secure architecture for digital twins in resource-constrained industrial systems.

Computing in Science Engineering, pages 1–12, 2025.

[9]    Elie Fute Tagne, Hugues M. Kamdjou, Adnen El Amraoui, and Armand Nzeukou. A lossless distributed data compression and aggregation methods for low resources wireless sensors platforms. Wireless Personal Communications, 128:621–643, 2022.

[10]    Yoshiteru Nagata, Daiki Kohama, Yoshiki Watanabe, Shin Katayama, Kenta Urano, Takuro Yonezawa, and Nobuo Kawaguchi. Semantipack: An efficient real-world data compressor using structural and semantic metadata. IEEE Access, 13:114159–114178, 2025.

[11]    Sifat Rezwan, Huanzhuo Wu, Juan A. Cabrera, Patrick Seeling, Martin Reisslein, and Frank H. P. Fitzek. Mlcxr+: Multilevel semantic compression for 3d immersion over 5g networks. IEEE Access, 13:164771–164786, 2025.

---

• Déposez votre candidature (Ouvre un nouvel onglet)