Phd for On-board Ai-based Fdir Architectures for - Immenstaad am Bodensee, Deutschland - Airbus

Airbus

Geprüftes Unternehmen

Immenstaad am Bodensee, Deutschland

vor 2 Wochen

Geschrieben von:

Lena Wagner

beBee Recruiter

Beschreibung

The implementation of AI-based Failure Detection, Isolation and Recovery (FDIR) is the core of this position.

Traditional FDIR methods based on models and signal processing might fail to detect faults that are below the pre-defined detection threshold and require additional hardware redundancy to detect faults.

The design of the FDIR includes a trade-off between the maximization of autonomy and mission availability on one side, and satellite design and validation complexity on the other side.

This latter being of particular importance since the FDIR is the last resource that guarantees the safety of the mission.

Hence, a robust validation concept is paramount especially for Ai-based FDIR concepts.

Tasks:

The objective of the PhD position is to define On-board AI-based FDIR architectures for satellite systems: from algorithm development to FDIR validation.

Following tasks have to be addressed:

Identify FDIR case studies for safety critical scenarios based on the previous Astrone project. Case study shall consider (but not necessarily be limited to)
Time series data and
Image data used by Vision-Based algorithms
Derive a baseline solution for the identified scenarios using traditional FDIR design techniques which are to be used as benchmark
Analyze how the diverse AItechniques known in literature compare when considering the identified case studies. Select and justify the most suitable approach through the process of rapid experimentation and prototyping.
Evaluate the suitability of using continuous learning AI methods to address the task of FDIR
Define a robust verification and validation strategy for the AI component, potentially based on blackbox, stochastic algorithm (e.g. a neural network).

This task shall include the analysis of:

Methods to formally ensure predictable and quantifiable behavior of the developed models
Validation criteria to ensure robustness of the developed models to different scenarios (for example, but not limited to inference time, precision, max false positive rate)
Scalability of the developed solution (training vs operational environment)
How the testing, verification and validation datasets should be generated and the type of telemetry data that can be used (only simulated telemetry data, only realtime telemetry data or combinations).

This task should also focus on:

The clear definition of the AI training process
Fidelity of the telemetry generated by the digital models (or satellite simulators) needed in order to effectively train the AI algorithms
Design a development framework covering the entire lifecycle of the selected AIbased FDIR solution (from algorithm development and training to deployment on a potential target hardware platform to continuous verification and validation)

Qualifications:

You have the following skills and qualifications:

Completed studies in Space Engineering, Data Science, Computer Science, Software Development or similar
Data Science, Development & Programming
Machine/deep learning frameworks (Scikitlearn, TensorFlow, Keras, PyTorch)
Experience in data exploration and interpretation, how to identify and model trends and deal with high dimensionality
Experience in at least one programming language (R, Python)
Experience with complex software environments and development frameworks (for example, acquire and process data from heterogeneous sources)
Good communication skills, a proactive mindset
Fluent English. Advanced Negotiation level in German is a plus.
Ability to define and manage work packages autonomously and work in close collaboration with other team members.
Experience in the following areas is seen as a strong asset:
background in descriptive statistics (clustering, dimension reduction, distribution fitting) and inferential/predictive statistics (Machine Learning such as Neural Networks, SVM, Random-Forest)
Experience with research methods (for example, graph theory, numerical optimization, surrogate modeling) to address challenges beyond data analysis, with pragmatic solutions as well (splines modeling, dichotomy optimization)

Not a 100% match? No worries Airbus supports your personal growth with customized development solutions.

This job requires an awareness of any potential compliance risks and a commitment to act with integrity, as the foundation for the Company's success, reputation and sustainable growth.

Company:

Airbus Defence and Space GmbH

Contract Type:

PHD, Research Contract / Doctorat, Contrat CIFRE / Doktorandenvertrag / Tesis doctoral

Experience Level:

Student / Etudiant / Student / Estudiante

Job Family:

At Airbus, we support you to work, connect and collaborate more easily and flexibly. Wherever possible, we foster flexible working arrangements to stimulate innovative thinking.