Die Universität Oldenburg engagiert sich für Gleichstellung, Diversität, Inklusion und die Vereinbarkeit von Familie und Beruf. Sie ist Unterzeichnerin der Charta der Vielfalt und seit 2004 als familiengerechte Hochschule zertifiziert. Wir begrüßen Bewerbungen von Menschen aller Nationalitäten, ethnischer und sozialer Herkünfte, Religionen, geschlechtlicher Identitäten, sexueller Orientierungen und Altersstufen.

« Zurück zur Übersicht

Stellenausschreibung / Job advertisement

The junior research group Adversarial Resilience Learning (Prof. Dr. Sebastian Lehnhoff), Division Energy Informatics, Department of Computing Science of the University of Oldenburg invites applications for a

Research Associate (m/f/d)
with focus on Explainable Deep Reinforcement Learning in Critical Infrastructures.

The position is full time (100%), pay rate E13 TV-L, for a period until 31st of July 2025, with the possibility of extension.

Adversarial Resilience Learning (ARL) is an opportunity to be part of the research in a new methodology for advanced deep reinforcement learning agents in critical national infrastructures (CNI), such as the power grid. In the ARL methodology, two (or more) competing agents (“attacker” and “defender”) compete for control of a CNI. Through this competition, the two agents learn robust strategies for a resilient control of the CNI. The research group develops an advanced architecture on top of it, featuring higher-order Deep Reinforcement Learning algorithms, extraction of learned strategies in a non-standard logic via methods of Explainable Reinforcement Learning, and learning from domain-expert knowledge.

Your research objective is to combine human knowledge, rule sets, and learned strategies in a hybrid agent architecture. Your research will allow the ARL agents to explain themselves in terms of behavioral rules encoded in Ternary Vector Lists (TVL), learn from human knowledge via TVLs, and distinguish between these rules and their own ability to learn and evolve. You will participate in a unique effort to combine basic research in a novel agent architecture with direct application to our society’s backbone system, most notably the modern power grid. Your research will be a part in the group’s goal to tackle current and future challenges to our critical national infrastructures, partnering with grid operators, an international cyber-security policy think tank, AI companies, and our partner universities Vanderbilt University and TU Delft.

Your Profile:

  • University degree in computer sciences; aptitude willingness to pursue a PhD that combines basic research in XRL with the practical application to CNIs, especially modern power grids
  • profound knowledge of at least two of: non-standard logics (esp. TVL and Boolean differential calculus), power grids, deep reinforcement learning, and explainable deep reinforcement learning
  • extensive programming experience with at least Python or C++, software engineering knowledge is a bonus
  • motivation and ability to publish and present scientific findings, represent the research group to international experts, and to collaborate with a diverse and international team of researchers
  • fluently spoken and written English.

What We Offer:

  • Fast-paced PhD in a prestigious, BMBF-funded junior research group along with publications and presentations on international conferences
  • Practical application of extensive basic research in tackling a pressing question for our society’s backbone systems
  • International network of partner universities as well as industry partners

The University of Oldenburg is an equal opportunities employer. According to § 21 para. 3 of the Legislation Governing Higher Education in Lower Saxony (NHG) preference shall be given to female candidates in cases of equal qualification. The same applies to persons with disabilities

Please send your application (letter of motivation, CV, copies of degrees, references) by email in a single pdf document with the keyword "ARL" to Christiane Großmann (). The closing date for applications is 2022-11-30.


(Stand: 17.11.2022)