UROP Project

Reinforcement Learning for Machine Resource Allocation



Jan Müller

Program Director UROP


+49 241 80-90299


Key Info

Basic Information

Project Offer-Number:
UROP International, RWTH UROP
Computer Science
Organisation unit:
RWTH Aachen University, Advanced Mining Technologies (AMT)
Language Skills:
English or German
Computer Skills:
- Some knowledge of numpy is a big plus;
- Work in a comfortable and creative environment within a focused team; - Aachen is a big center of engineering; - Aachen is located in the three country corner, lots to see and travel; - Students get a regional train pass to travel around; - Institute is located right next to the city center, everything is close by
Univ.-Prof. Dr.-Ing. Elisabeth Clausen


The interest in Reinforcement Learning (RL) has recently surged. RL is the computational approach to learning from interaction. A RL agent is mapping situations it encounters within a given environment to actions that lead to the most reward. The agent does not always know the environment beforehand and therefore has to discover what actions yield the most reward. We at the Advanced Mining Technology (AMT) Institute are committed to providing technologies and system solutions for harsh environments. Open Pit-Mining is one such harsh environment, in which a multitude of mining machines, ranging from dump trucks to excavators, have to co-operate to extract mineral resources. Aim of this project is to develop a reinforcement learning model that can allocate mining tasks to machine resources so that mineral resource extraction is conducted as efficiently as possible. As resource efficiency and social responsibility are our vision and misson, this project brings our core values to life.


- Set-Up and refine a simulation environment - Choose between different reinforcement learning algorithms/models - Implement reinforcement learning models - Evaluate test results


- Intermediate Python Skills (must) - Research oriented mindset (must) - Strong problem solving skills (must) - Know-how in stochastic modelling (optional)