Internal Research Fellow (PostDoc) in Advanced onboard processing architectures for space systems

Our team and mission

The Data Handling Section provides functional support to ESA projects and carries out technological research and development (R&D) concerning turn-key on-board hardware data handling solutions, with an emphasis on:

• platform and payload data handling architectures and their building blocks (equipment/units, modules and key components);
• units such as onboard computers, mass memories, remote terminals and instrument control units*;
• digital and analogue signal processing electronics for payload/platform functions;
• front-end acquisition and processing chain electronics*;
• onboard data transfer interfaces, buses and associated protocols (high, medium and low speed);
• platform data handling functions related to security, data authentication, encryption/decryption and compression;
• use of microelectronics devices (including COTS);
• implementation, inference, verification and validation of algorithms** on processing hardware platforms for space applications* in close collaboration with other discipline experts (software, microelectronics and applications engineers).

* except for RF payloads.
** including artificial intelligence and machine learning algorithms.

You are encouraged to visit the ESA website: https://www.esa.int/

Field(s) of activity/research for the traineeship

The main objective of the proposed Research Fellowship consists in carrying out research in the field of advanced onboard data processing architectures for space systems. In particular, one of the two following topics might be considered as the objective of this Research Fellowship:

(a)  Assessing the performance and fault tolerance of neuromorphic hardware;
(b)  Designing and developing one or more machine learning (ML) and artificial intelligence (AI) algorithms to support and enhance the autonomous event reaction management part of the fault detection, isolation and recovery (FDIR) software. 

(a) Assessing the performance and fault tolerance of neuromorphic hardware

The main objective of this research topic will be the assessment of the performance and fault tolerance of neuromorphic hardware.  In particular, the focus will be on the benchmarking of its effectiveness when processing data from conventional, frame-based sensors in comparison to event-based sensors, which are more commonly associated with neuromorphic computing. This work will support mission design decisions regarding the suitability of neuromorphic systems for different types of onboard sensing and processing tasks in the space environment.

Your research will be guided by your own expert judgement and insight into current trends in AI hardware acceleration and neuromorphic computing, while also aligning with the hosting team’s strategic priorities in onboard autonomy, radiation robustness, and efficient edge processing.

Scientifically, you will in particular:

• propose and conduct comparative benchmarking of neuromorphic hardware using both event-based sensor data and conventional sensor data (such as optical or hyperspectral imagers typical of space missions). Metrics will include energy efficiency, latency, throughput, fault tolerance and inference accuracy.
• analyse the suitability of neuromorphic computing for various space-relevant processing tasks—such as visual navigation, object detection, anomaly identification, and compression—when using conventional versus event-based sensing modalities.
• identify application domains where neuromorphic systems provide a performance or efficiency advantage, and determine scenarios where conventional AI accelerators (such as embedded GPUs or FPGA-based accelerators) remain more appropriate due to data characteristics, or hardware or algorithmic constraints.
• investigate the fault tolerance properties of neuromorphic processors in radiation-prone environments, including the effects of single-event upsets (SEUs), latch-ups, and the total ionising dose on spiking neural network performance.
• develop and test fault mitigation strategies, such as spike-based redundancy, reconfigurable neural routing, noise-aware learning and population coding, to improve neuromorphic system resilience during in-orbit operation.
• contribute to the establishment of a methodology for evaluating neuromorphic versus traditional AI acceleration approaches across different sensing and fault scenarios, thereby informing hardware and architecture trade-offs for future missions.

(b) Designing and developing ML and AI algorithms to enhance FDIR

The main objective of this research topic will be the assessment of the fault detection, isolation and recovery (FDIR) functionality and performance, mainly focusing on the event recovery part. The goal will be to design and develop one or more machine learning and artificial intelligence-based algorithms to support the autonomous event reaction management part of the FDIR software. This effort aims to significantly advance future FDIR implementation. While substantial progress has been made in applying ML-based algorithms for failure detection, an onboard solution for autonomous event reaction remains largely unfulfilled. This work has the potential to introduce greater onboard autonomy, thereby enhancing mission reliability and the satellite's nominal operational capabilities.

Scientifically, you will in particular: 

• identify the primary avionics and data handling signals that will inform the development of a potential general-purpose algorithm adaptable to various missions, such as for Earth observation and science, based on insights from the classical FDIR approach. This selection will be foundational in defining the inputs for evaluating the event reaction algorithm.
• design and develop ML and AI algorithms to support failure event reactions, considering state-of-the-art algorithms and classical FDIR approaches. Create algorithms that closely align with actual flight applications, aiming to enhance performance. 
  Validate and test these solutions with respect to the classical methods.
• choose an appropriate data processing board, optimise the existing software that has been developed, tested, and validated, and then integrate the software on the selected board to assess its performance and derive hardware benchmarks. 
• develop a methodology and guidelines for the effective development and implementation of AI event reaction algorithms to support future missions. 

Technical competencies

Behavioural competencies

Result Orientation
Operational Efficiency
Fostering Cooperation
Relationship Management
Continuous Improvement
Forward Thinking

For more information, please refer to ESA Core Behavioural Competencies guidebook

Education

You should have recently completed, or be close to completion of a PhD in a related technical or scientific discipline. Preference will be given to applications submitted by candidates within five years of receiving their PhD.

Additional requirements

In addition to your CV and your motivation letter, please prepare a research proposal of no more than 5 pages. This proposal should be uploaded to the "additional documents" field of the "application information" section.

Depending on the topic selected for the research, the following technical competencies would be considered assets: 

For research topic (a): 

• Knowledge and understanding of event-based sensors (cameras).
• Knowledge of neuromorphic hardware architectures
• Knowledge of AI-accelerator architectures
• Understanding of radiation effects in electronics

For research topic (b):

• Good knowledge and understanding of ML and AI architectures for hardware and software implementation
• Understanding of FDIR concepts, as applied in on-board space systems
• Understanding of avionics and onboard data handling architectures

You should also have good interpersonal and communication skills and should be able to work in a multi-cultural environment, both independently and as part of a team. Your motivation, overall professional perspective and career goals will also be explored during the later stages of the selection process.

The working languages of the Agency are English and French. A good knowledge of one of these is required. Knowledge of another Member State language would be an asset.