Problem Description:

Production planning plays a crucial role in improving efficiency and resource utilization in electronics production. Many production planning problems can be formulated as combinatorial optimization problems and solved using specialized solvers. However, in real-world industrial settings, the vast amount of data leads to extremely high-dimensional optimization problems, which are significant challenges for classical solvers. As a result, developing solvers handling high-dimensional optimization problems has become a popular research topic. In recent years, applying artificial intelligence algorithms, particularly reinforcement learning, to solve complex, high-dimensional combinatorial optimization problems has attracted considerable attention.

Some studies showed that reinforcement learning has potential in solving combinatorial optimization problems. However, it is not that reliable when dealing with large-scale industrial data. Traditional RL methods often struggle with extremely slow convergence and sometimes get stuck in local minima, making it hard to apply them in real-world industrial scenarios.

This thesis aims to explore new ways to improve the efficiency of reinforcement learning algorithms, so they can better handle high-dimensional optimization problems.

Research topics and Workplan:

– Implementing classical reinforcement learning algorithms using a current popular software framework to solve a pre-defined small-scale industrial optimization problem.
– Explore methods to integrate traditional reinforcement learning algorithms with stochastic optimization algorithms.
– Benchmarking the classical reinforcement learning algorithms against your methods.
– (Optional) Exploring the possibilities of quantum reinforcement learning for solving the pre-defined industrial optimization problem.

Your abilities:
– Programming skills in Python are necessary.
– Prior knowledge or experience on reinforcement learning and reinforcement learning framework (Pytorch, Tensorflow, Gym…).
– Good English or German skills.
– You can begin with your BA/PA/MA as soon as possible.

You can contact us via the emails below. Please make sure to attach your transcript and resume when applying, and include a brief motivation in the email. Your motivation must be relevant to the topics described above. We look forward to your participation in our research.

Kategorien:

Kontakt: