Chance-Constrained Control with Lexicographic Deep Reinforcement Learning

📖 Description

THIS IS A PREPRINT VERSION. IF YOU FOUND THIS READ ING USEFUL FOR YOUR RESEARCH PLEASE CITE THE PUBLISHED VERSION https://ieeexplore.ieee.org/abstract/document/9031720 DOI: https://doi.org/10.1109/LCSYS.2020.2979635 BIB: @article{Giuseppi2020, doi = {10.1109/lcsys.2020.2979635}, year = {2020}, author = {Alessandro Giuseppi and Antonio Piet rabissa}, title = {Chance -Constrained Control with Lexicographic Deep Reinforcement L earning}, journal = {{IEEE} Control Systems Letters}} Abstract ?This paper proposes a lexicographic Deep Reinforcement Learning (DeepRL )-based approach to chance -constrained Markov Decision Processes, in which the controller seeks to ensure that the probability of satisfying the constraint is above a given threshold. Standard DeepRL approaches require i) the constraints to be included as additional weighted terms in the cost function, in a multi -objective fashion, and ii) the tuning of the introduced weights during the training phase of th...

📊 Game Impacts

💭 Reactions

🔗 Sources

• [1] https://arxiv.org/abs/2010.09468