198-Autonomous time scale in a swarm of satellites using data fusion

198-Autonomous time scale in a swarm of satellites using data fusion

  • Contract :Ph.D.
  • Duration :36 months
  • Working time :Full-time
  • Experience :Entry Level
  • Education level :Master’s Degree, MA/MS/MSc

Your mission at CNES :

A satellite swarm is a specific type of constellation that aims at replacing large and complex satellites by a large number of simple and small elements. In some cases, a satellite swarm can provide services that cannot be obtained by a single satellite: the NOIRE study (Nanosatellites for a Radio Interferometer Observatory in Space [1]), performed by CNES and Observatoire de Paris – PSL, gives an example of such a service, where a large number of nanosatellites is used to observe fossil radio signals with frequencies inferior to 100 MHz. The method typically used to observe such signals consists in applying interferometric methods to pairs of continent-sized antenna arrays. However, in the case of the signals considered in NOIRE, because of the presence of both the ionosphere and the Earth radio pollution, the antenna array has to be placed in space, in a point where it is protected from Earth emissions. The closest point is the far side of the Moon.
The instrument that could be used in NOIRE has to perform mostly autonomously. One of the functions required by the scientific mission is the synchronisation of its elements with a precision of a nanosecond. This requires the creation of an autonomous time scale, based on the clocks available on each satellite of the swarm. According to [2], “a time scale is considered as one of the coordinates of a four-dimensional space-time reference system. In practice, the generation of an accurate, stable, and reliable time scale calls for the design of an algorithm able to efficiently combine timing data from an ensemble of clocks”. In the case of NOIRE, the main design constraints will be real-time access and robustness to anomalies and clock dysfunctions.
Main purpose:
Generate a stable time scale from the clocks of a swarm of nanosatellites, adapted to the requirements of a specific scientific mission.
Detailed objectives:
The first step consists in performing an analysis of the scientific objectives of the NOIRE mission, to derive the set of performance requirements for the developed time scale, i.e., which of the usual performance indicators (reliability, stability, accuracy and delay of access) are actually crucial and which indicators have less significance in this context.
The next step will be to analyse the existing bibliography to determine the general architecture of the proposed algorithm for designing an appropriate time scale. Many time scales of reference such as the Temps Atomique Français TA(F) [3] rely on ARIMA processes. However, the Kalman filter is often described as a potential evolution (e.g., the Coordinated Universal Time [4] uses the ALGOS(BIPM) algorithm but the potential use of a Kalman filter has been studied in [5]). According to [6], the use of a Kalman filter may cause a degradation in short term stability. Another problem noted in [6] is the unlimited growth of some elements of the state covariance matrix due to the unobservability of certain parameters. Nevertheless, Kalman filter-based time scales have been introduced to deal with situations very close to the NOIRE mission, such as [7] which designs an autonomous clock ensemble for a GNSS constellation.
After defining the general architecture and performance criteria, the PhD candidate will have to design more precisely the proposed time scale. In practical terms, this consists in computing a weighted average of all the free-running clocks of the swarm to obtain a synthetic clock that is more stable in the long term than each of its individual components. The PhD candidate will therefore have to define the weight computation algorithm that is best suited to the mission’s scientific purposes, and the type of access method that provides the clock readings (from one-way TDMA to full-duplex CDMA). In particular, the following issues will have to be considered:
Computational power: the time scale needs to be computed and available on board, by the nanosatellites themselves, i.e., the algorithm needs to be distributed and/or very light.
Clock technology: what kind of clock may be used in a nanosatellite? An adapted noise model will have to be defined. The reliability may not be very high: the probability of anomalies and dysfunctions may be high. However, Chip Scale Atomic Clock standards are now becoming available. The impact of clock stability will also depend on the type of access method: if TDMA is used, the collection of clock readings may take some time and the reliability of the resulting time scale will depend on the stability of each clock.
Anomaly detection: since the time scale needs to be not only available but also reliable in real time, any clock malfunctions needs to be detected as fast as possible. Many anomaly detection methods studied in the machine learning community could be used for this detection. An alternative is to use robust estimation methods that are not very sensitive to clock problems. Since malfunctions will still be (hopefully) rare events, sparse estimation may finally be considered.
Research Area for this PhD topic:
Data fusion
Signal processing
Robust estimation
Sparse estimation
Machine learning
[1] B. Cecconi, M. Dekkali, C. Briand, B. Segret, J. N. Girard, A. Laurens and P. Gélard, “NOIRE study report: Towards a low frequency radio interferometer in space.,” in Proc. IEEE Aerospace Conference, Big Sky, Montana (USA), 2018.
[2] C. Thomas, P. Wolf et P. Tavella, «Time Scales,» Bureau International des Poids et Mesures, Sèvres, 1994.
[3] C. Andreucci, «A new algorithm for the French atomic time scale,» Metrologia, vol. 37, n°1, 2000.
[4] G. Panfilo et F. Arias, «The Coordinated Universal Time (UTC),» Metrologia, vol. 56, n°042001, 2019.
[5] P. Federica et G. Panfilo, «A new approach to UTC calculation by means of the Kalman Filter,» Metrologia, vol. 53, n°1185, 2016.
[6] M. Weiss et T. Weissert, «AT2, A New Time Scale Algorithm: AT1 Plus Frequency Variance,» Metrologia, vol. 28, n° 65, 1991.
[7] M. Coleman et R. Beard, «Autonomous clock ensemble algorithm for GNSS applications,» Navigation-US, vol. 67, pp. 333-346, 2020.

For more information, contact from TéSA

Candidate profile searched:

Master in data fusion, signal processing, statistics, and machine learning

We suggest you to contact first the PhD supervisor about the topics and the co-financial partner (found by the lab !). Then, prepare a resumé, a recent transcript and a reference letter from you M2 supervisor/ engineering school director and you will be ready to apply online !

CNES will inform about the status of your application in mid-June.

More details on CNES website :

  • Our address

    7 Boulevard de la Gare, 31500 Toulouse, France
    Would you like to calculate the access time? Click here