Les commentaires sont clos.

PhD position: Artificial intelligence for telecommunications in adverse environments

27 Septembre 2022

Catégorie : Doctorant

CIFRE PhD position at Schlumberger in artifical intelligence and telecommunication



Our story begins with what it truly means to be a technology innovator. It stems from a common sense of purpose that unites the people of Schlumberger who, representing more than 160 nationalities, provide leading digital solutions and deploy ground-breaking technologies to enable performance and sustainability that are crucial for the global energy industry. With expertise in more than 120 countries, we partner with customers in close collaboration to create industry-changing technologies that unlock cleaner, safer access to energy for every community— including those we live and work in.

Schlumberger Riboud Product Center (SRPC)is the largest Schlumberger technology and development center in Europe. Around 800 scientists, engineers, and technicians, of more than 50 nationalities, design and manufacture equipment and systems for our oilfield services worldwide. Based in Clamart close to Paris, SRPC teams form a center of excellence for research and development of breakthrough technologies. SRPC is recognized globally for its expertise in:

  • Artificial intelligence
  • High-temperature electronics
  • Mechanical systems for extreme conditions
  • Physics of sensors and measurements
  • Software development
  • Applied mathematics
  • Geophysics and geology
  • Energy engineering

Our strength comes from our passion for innovation and our multicultural population.

The PhD will take place mainly atSRPC with regular stays at the Mathematical and Electrical Engineering department of IMT Atlantique (



Schlumberger develops tools for oil exploration, production optimization and maintenance operations. These tools provide measurement resources whose critical information must be transmitted to the end user.

The efficiency and reliability of the means of transmitting critical information is at the heart of the problems of making the physical layer reliable. Schlumberger uses different types of telecommunication systems operating on various physical layers:

  • Communication in the drilling mud or by long-distance electric field for the transmission of geophysical measurements during drilling
  • Communication through metallic drill pipe for transmission of measurements during production testing
  • Underwater acoustic communication for the transmission of inspection and maintenance measurements of subsea equipments
  • Communication through long electrical cables for transmission of measurements related to hydrocarbon reservoir evaluation.

Schlumberger has developed specific expertise in the design of these different telecommunications means [1-3]. Compared to the physical layer used in terrestrial mobile telephony, the physical layer of the communication systems developed at Schlumberger is very severe. The operational environment is characterized by an extremely high noise level (several hundred times more energetic than the signal), a very selective communication channel (up to 90% of the frequency band can be lost), and a constant fluctuation of the parameters over time. The reliability of these systems relies to a large extent on a detailed understanding of the underlying channel and noise models.

In the general case, the detailed understanding of the physics and the associated modeling efforts can be extremely difficult and costly. For this reason, Schlumberger wants to facilitate the design stages by learning the necessary information directly from raw data using machine learning/artificial intelligence techniques.



The PhD work will be organized in four distinct phases aimed at evaluating the application of artificial intelligence algorithms applied to the physical layers.

In phase 1, a comprehensive literature review of artificial intelligence techniques applied to the physical layers of interest will be conducted. A channel and noise model will then be designed to create a large database with real or simulated test and training examples.

Phase 2 involves the development of reinforcement learning algorithms for automatic decision making during operational changes in communication receivers. In the general case, a change in channel conditions or noise signatures requires the calibration of hyper-parameters to optimize the operating point of the modem. We will focus here on the development of reinforcement learning algorithms that will aim at learning automatic optimization strategies for these hyper-parameters.

Phase 3 concerns the development of modular telecommunication algorithms. While maintaining a tiered telecommunication architecture of the type: frame detection, noise cancellation, spatial combination, equalization and decision, this phase will propose improvements to standard processing blocks based on simplified models. The objective is to specialize the models to realistic noise and channel conditions using complex statistical learning architectures. For this purpose, the examples generated during Phase 1 will be particularly useful for learning the statistical models and specializing the processing blocks for representative noise and channel conditions. In addition, we will be particularly interested in the computational cost issues of equalization and noise cancellation algorithms.

Phase 4 will investigate the consideration of imperfections in the analog-to-digital and digital-to-analog communications chains. In the context of the physical layers of interest, the signal sources (e.g. electromagnetic antennas, piezoelectric transducers,...), the amplification stages (e.g. class B or D amplifiers) as well as the reception elements (e.g. antennas, accelerometers, piezoelectric sensors) introduce non-linearities in the communication chain. The objective of this phase will be to evaluate learning techniques aimed at l correcting the defects of the transmission chains



  • Master’s Degree or Engineering degree with a solid background in either: applied mathematics, signal processing, telecommunications, or machine learning
  • Working knowledge of relevant software, C++/Python/Matlab. Experience in software defined radio would be a plus (GNU-radio, Seeedstudio, NESDR…).
  • Working knowledge of relevant machine learning frameworks (Tensorflow, Pytorch, Theano…)



Applications should include a CV, a statement of interests and motivations, a reference letter, as well as academic transcripts (Master level) and must be sent to Arnaud Jarrot ( and François-Xavier Socheleau (



[1] A. Jarrot, A. Gelman and J. Kusuma, "Wireless Digital Communication Technologies for Drilling: Communication in the Bits/s Regime," in IEEE Signal Processing Magazine, vol. 35, no. 2, pp. 112-120, March 2018, doi: 10.1109/MSP.2017.2781288.

[2] J. Kusuma, A. Jarrot, A. Gelman, A. Croux and G. Choi, "Pragmatic Performance Optimization of a Multichannel DFE System for a Wideband 100 kbps 1-km Subsea Acoustic Modem," 2018 Fourth Underwater Communications and Networking Conference (UComms), 2018, pp. 1-4, doi: 10.1109/UComms.2018.8493222.

[3] A. Jarrot et al., "High-speed underwater acoustic communication for multi-agent supervised autonomy," 2021 Fifth Underwater Communications and Networking Conference (UComms), 2021, pp. 1-4, doi: 10.1109/UComms50339.2021.9598036.

[4] A. Pottier, F-X. Socheleau, C. Laot, "Quality-of-Service Satisfaction Games for Noncooperative Underwater Acoustic Communications", IEEE Access, Topic : Underwater Wireless Communications and Networking, Vol. 6, Apr. 2018

[5] A. Pottier, F-X. Socheleau, C. Laot, “Robust Noncooperative Spectrum Sharing Game in Underwater Acoustic Interference Channels", IEEE Journal of Oceanic Engineering, 2017

[6] Pottier, A., Mitchell, P. D., Socheleau, F. X., & Laot, C. Q-learning based adaptive channel selection for underwater sensor networks. In 2018 Fourth Underwater Communications and Networking Conference (UComms) (pp. 1-5). IEEE.

[7] Zhang, Y., Li, J., Zakharov, Y. V., Li, J., Li, Y., Lin, C., & Li, X. (2019). Deep learning based single carrier communications over time-varying underwater acoustic channel. IEEE access, 7, 38420-38430.

[8] Aoudia, F. A., & Hoydis, J. (2018, October). End-to-end learning of communications systems without a channel model. In 2018 52nd Asilomar Conference on Signals, Systems, and Computers (pp. 298-303). IEEE.

[9] Chen, M., Challita, U., Saad, W., Yin, C., & Debbah, M. (2019). Artificial neural networks-based machine learning for wireless networks: A tutorial. IEEE Communications Surveys & Tutorials, 21(4), 3039-3071.

[10] Shlezinger, N., Farsad, N., Eldar, Y. C., & Goldsmith, A. J. (2020). ViterbiNet: A deep learning based Viterbi algorithm for symbol detection. IEEE Transactions on Wireless Communications, 19(5), 3319-3331.