Reinforcement Learning

What is the optimal action that you should take in a specific situation? This question is not only raised in our everyday life, it is also relevant to every control system. In this research, I address this question with a machine learning methodology called reinforcement learning. Furthermore, I work on several real world use cases such as:

* Identifying the optimal reorder policy for a set of warehouses to minimize the overall logistics costs under the assumption of stochastic demand and stochastic lead times.
* Finding the elevator group control policy that minimises the travel time of each passenger in Switzerlands highest building.
* Finding the optimal building automation control to minimize energy consumption and maximize the comfort.

If you are interested in diving deeper into this topic, I would recommend you to (1) have a coffee chat with me and (2) go through the following literature/videos: https://www.youtube.com/watch?v=VMp6pq6_QjI and https://www.youtube.com/watch?v=WXuK6gekU1Y provide you with an intuitive understanding of what reinforcement learning “is doing”. https://www.youtube.com/watch?v=93M1l_nrhpQ goes deeper and is also highlighting the underlying mathematics. Furthermore, https://www.youtube.com/watch?v=W_9kcQmaWjo covers the topic multi-agent reinforcement learning, which plays an essential part in this research.

Email: patric.hammler@roche.com

Related Literature:
[1] Supply chain optimization: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3302881
[2] Elevator group control: https://ieeexplore.ieee.org/abstract/document/8998335
[3] Building automation: https://dl.acm.org/doi/abs/10.1145/3360322.3360857