PhD student for Artificial intelligence to improve MOX chemo-resistive sensor networks (electronic noses) - Fixed term contract - 3 years

ABOUT TELECOM SUDPARIS

Telecom SudParis is a public graduate school for engineering, which has been recognized on the highest level in the domain of digital technology. The quality of its courses is founded on the scientific excellence of its faculty and on teaching techniques that emphasize project management, innovation and intercultural understanding. Telecom SudParis is part of the Institut Mines-Telecom, the number one group of engineering schools in France, under the supervision of the Minister for Industry. Telecom SudParis with Ecole Polytechnique, ENSTA Paris, ENSAE Paris, ENPC and Telecom Paris are co-founders of the Institut Polytechnique de Paris, an institute of Science and Technology with an international vocation.

Its assets include: a personalized course, varied opportunities, the no.3 incubator in France, an ICT research center, an international campus shared with Institut Mines-Telecom Business School and over 60 student societies and clubs.

CONTEXT :

Air and its impact on public health are a growing concern. Although some criteria are established for outdoor environments, no precise system is currently deployed in indoor environments. The challenge lies in measuring the concentrations of volatile organic compounds (VOCs) in indoor air, where different gas mixtures can be present at the same time, which considerably complicates the situation.

The democratization of portable sensor-based measurement systems represents an opportunity to understand the consequences of air quality on individual health.

Metal oxide gas sensors (MOX sensors) dominate the market for ready-to-use gas sensors thanks to their miniaturization, low cost and availability. However, they are not generally used individually to detect a specific gas, as they are sensitive to many parameters, including a lack of sensitivity.

The solution studied in most applications is to group these sensors into clusters (sometimes called electronic noses) containing different models of MOX sensors capable of measuring various gas species with different sensitivity levels. The question arises to find which combination of individual sensors to consider to obtain an electronic nose allowing a precise identification of the specific gases of a given application. This is one of the concerns of the AMUSENS European project to which this thesis will contribute.

Recent studies have highlighted several challenges related to the use of MOX sensors in natural environments. Among these are their lack of sensitivity, their dependence on temperature and humidity variations, as well as their drift (the gradual and unpredictable variation of their response over time, even in the presence of identical concentrations).

Although some solutions have already been proposed to mitigate these limitations, the design of efficient multi-sensor devices (electronic noses) capable of accurately identifying VOCs in indoor environments remains an open challenge.

MISSIONS :

This thesis focuses on the following problems:

1) In the AMUSENS project, we will develop AI algorithms to identify the optimal combination of MOX sensors to efficiently detect pre-selected volatile organic compounds (VOCs).

2) To overcome the drift problem [Dennler et al.-2022] and other inconsistencies in MOX sensors when used in uncontrolled conditions, a possible approach is to represent these signals in “contextual” or semantic spaces. This idea generalizes the concept of word embeddings, where non-linear functions of conditional probabilities are considered as (pseudo) distances between words, or any categorical symbol that can coexist in a given context. Furthermore, in artificial neural networks, especially those of processing architectures called Transformers, ambiguous contextual information already embedded in these spaces can be partially removed using a message-passing approach known as Attention layers.

3) Another scientific concern regarding MOX sensor clusters is their ability to represent different classes of signals. In many experiments with MOX sensor-based E-noses, it has been observed that the signals seem to cover a relatively small space with 3 or 4 degrees of freedom, a characteristic referred to as the intrinsic dimension (ID). This persistently low ID value should be a constraint for the classification of volatile compounds with MOX-based E-noses.

References:

N. Dennler, S. Rastogi, J. Fonollosa, A. Van Schaik et M. Schmuker, « Drift in a popular metal oxide sensor dataset reveals limitations for gas classification benchmarks » , Sensors and Actuators B: Chemical, vol. 361, p. 131668, 2022.

ACTIVITIES :

The tasks to be carried out as part of this thesis are as follows:

- Produce a bibliography on the subject

- Study, simulate and develop algorithms responding to the issues set out in the previous section (Missions) exploiting existing and new data produced in the AMUSENS project

- Write publications for conferences and journals in the field

- Write a thesis dissertation

Level of training and / or experience required:

-Master degree or equivalent

- Master Degree in Research in the fields of Data Science, AI, Pattern Recognition, Signal and Data Processing

Essential skills, knowledge and experience:

- Deep learning methods

- Data processing and classification methods

- Basics in signal processing and embedded electronics

- Knowledge of the field of MOX gas sensors highly appreciated

- Languages ​​Python, Matlab, C, Unix,….

Advantageous skills, knowledge and experience:

- Desirable knowledge in the field of air quality

- Knowledge of scientific library software specific to Machine Learning, Signal and Data Processing (Pytorch, KERAS, etc.)

Abilities and skills:

- Fluency in English (written and spoken) and French (very apreciated)

- Teamwork ability

- Analytical skills and rigor

- Ability to tackle cross-disciplinary scientific subjects (AI, signal processing and electronics and MOX sensors)

- Ability to work in an international scientific environment, the thesis being carried out within the framework of a European project (AMUSENS project)

Further information and application

• Application deadline: April 09, 2025
• Type of contract: 3-years doctoral fixed-term contract
• Job location: Evry-Courcouronnes (91) - regular travel to the Palaiseau site (91) required.
• Positions offered for recruitment are open to all, with accommodations available on request for candidates with disabilities.
• Working conditions : Teleworking possible, on-site restaurant and cafeteria, accessibility by public transport (with employer's contribution) or close to main roads, staff association and sports association on campus.
• Contact : Jérome BOUDY - jerome.boudy@telecom-sudparis.eu