Applications
A nice direction for the course is to apply the algorithm in your paper to a new application. We will focus on two types of applications with real-world relevance: 1) energy management, and 2) quantum problems.
Gymnasium environment: Make sure that your application has a readily available Gymnasium environment. If the codebase of your paper also uses Gym, then you can simply swap out the environments.
Match paper and application: If you want to study a particular application, make sure the paper you choose matches with the application (e.g., you cannot do multi-agent RL in a single-agent problem). We indicate these matches below.
Below you find relevant Gymnasium packages!
RL for Sustainable Energy
Building energy management
Beobench: Codebase. Documentation. (white paper). This package integrates several packages for building energy management simulation:
BOPTEST: Codebase. Documentation.
SinerGym: Codebase. Documentation. (white paper)
Energym: Codebase. Documentation.
Possible algorithms: model-free RL, model-based RL, (transformers).
Electricity distribution networks
CityLearn: Codebase. Documentation. (white paper) ----- Multi-agent RL paper with code
Gym-ANM: Codebase. Documentation.
PowerGridworld: Codebase. Documentation. (white paper)
Possible algorithms: multi-agent RL (with multiple buildings), model-free RL.
RL for Quantum
Quantum games
Quantum Tiq-Taq-Toe: Example play. Code (you need to write a Gymnasium wrapper around this code)
Possible algorithms: model-free RL, (model-based RL, transformers).
RL for Astronomy
Telescope Imaging
Adaptive Optics. Code. Documentation/explanation. Paper.
Possible algorithms: model-free RL, (model-based RL, transformers).