PhD Student

Job description

PhD position in Safe and Goal-directed Adaptive Control
Research Group active on system dynamics and control within the Faculty of Engineering, Ghent University; is looking for outstanding applicants for a doctoral position.

About us
Our research group works on system dynamics and control; and more specifically on the modelling, control and design of mechatronic systems, robots and energy systems. We are part of the department of Electromechanical, Systems and Metal Engineering within the Faculty of Engineering of Ghent University (www.ugent.be/ea/emsme/en). Ghent University is a top 100 university worldwide and one of the major universities in Belgium, with more than 44000 students and 15000 staff members. Our campus is situated in in Ghent, a lively city at the heart of Europe (visit.gent.be/en/HOME). Our research group is also associated to Flanders Make (www.flandersmake.be/en), a non-profit organization, funding precompetitive research for the betterment of the Flemish manufacturing industry. The candidate will be directly embedded in an international research group (https://dynamics.ugent.be/), working together as a team and will have the possibility to collaborate with many other people active at Ghent University and within Flanders.

Job description
You will work on the Flemish AI Research program and the CTRLxAI, a Strategic Basic Research project funded by the Flemish scientific research fund (FWO-Flanders).
At the intersection of these two projects, we aim to stimulate cross fertilization between more than 150 years of successful control systems theory and recent successes of (deep) Reinforcement Learning (RL). RL and machine learning approaches in general have shown their value in virtually all areas of science and engineering. The goal is to develop a new family of hybrid adaptive control methods seeking 4 objectives
(1) improved range of operation and performance,
(2) improved data/sample efficiency
(3) guaranteed safety (or stability)
(4) guaranteed transparency in action selection.

Where traditional adaptive controllers usually excel in (2-4), RL methods excel at (1). This can be explained by the way these respective paradigms are parameterized and how they adapt the closed loop control or policy. Traditional adaptive control methods are usually indirect: They rely upon models that consist of parameters that on their turn are updated based on parameter estimation approaches. As a counterpart, RL is a ‘direct’ adaptive control approach that directly aims adapts actions based on the immediate and/or long-term performance improvement (as defined/quantified in whatever way seems relevant to the application). Since RL does not put any restriction on the policy parameterization, in theory the optimal policy can be identified at the cost of a high data dependency and a lack of any stability guarantees. Traditional methods on the other hand restrict the representational power of the policy by incorporating a lot of prior knowledge, gaining insight and stability at the cost of a decrease in performance and large engineering effort.
Your PhD research will specifically focus on the development of an adaptive control architecture that relies on traditional control methods (PID, MPC, …) to parameterize the policy and on Reinforcement Leaning to adapt the parameters in a goal-directed manner. In a second phase, this framework will be extended to develop an adaptive architecture with dual properties, actively balancing the exploration/exploitation trade-off. You will apply and validate your algorithms on real-world dynamical systems.
 
The candidate will be expected to

• Perform high-quality research and strive towards successful project execution.
• Develop software tools (Python) for probabilistic optimization & identification.
• Present research at conferences and in journals.
• Cooperate with researchers active within the research group and outside.
• Contribute to the teaching related to modelling, control and optimization of dynamical systems.
  
  

WHAT WE CAN OFFER YOU

• We offer a full-time position as a doctoral fellow, consisting of an initial period of 12 months, which - after a positive evaluation, will be extended to a total maximum of 48 months.
• Your contract will start on 1st September 2024 (indicative).
• The fellowship amount is 100% of the net salary of an AAP member in equal family circumstances. The individual fellowship amount is determined by the Department of Personnel and Organization based on family status and seniority. A grant that meets the conditions and criteria of the regulations for doctoral fellowships is considered free of personal income tax. Click here for more information about our salary scales
• All Ghent University staff members enjoy a number of benefits, such as a wide range of training and education opportunities, 36 days of holiday leave (on an annual basis for a full-time job) supplemented by annual fixed bridge days, bicycle allowance and eco vouchers. Click here for a complete overview of all the staff benefits (in Dutch).
• Access to state-of-the-art tools and facilities, a network of Flemish companies active in the manufacturing industry, and the possibility to collaborate with other research groups.
• The time to apply and improve your knowledge and skills on state-of-the-art machine learning, (probabilistic) modelling, system identification and numerical optimization.

Job profile

A background in modelling, control & numerical optimization methods and experience with system identification concepts is an asset. As a person you are quick-witted, learn fast and program faster. There are many ideas pending implementation and many more to explore. You feel at ease doing either.
Hard skills

• You hold a M.Sc. in electromechanical engineering or related engineering fields such as control & automation.
• You have proven experience with numerical optimization methods in system design (and machine learning).
• You have experience in probabilistic methods (and machine learning).
• You have proven experience in Python.

Soft skills

• You have a team player mindset, a strong personality and work in a result-oriented manner.
• You are creative and willing to work in a multidisciplinary context.
• You are proficient in oral and written English and have strong communication skills.
• You are willing to extend your network and able to talk on technical matters.

How to apply

Interested?
Send your CV, containing 1 or more references and a brief motivation letter to Guillaume Crevecoeur (Guillaume.Crevecoeur@ugent.be) including ‘CTRLxAI PHD’ in the email subject before 7th July 2024. If you pass the pre-selection, you will receive further instructions on the selection process and will be invited for an online job interview.