A Deep Reinforcement Learning Architecture for Multi-stage Optimal Control

doi:doi:10.57702/y6obe121

A Deep Reinforcement Learning Architecture for Multi-stage Optimal Control

Deep reinforcement learning for high dimensional, hierarchical control tasks usually requires the use of complex neural networks as functional approximators, which can lead to inefﬁciency, instability and even divergence in the training process. Here, we introduce stacked deep Q learning (SDQL), a ﬂexible modularized deep reinforcement learning architecture, that can enable ﬁnding of optimal control policy of control tasks consisting of multiple linear stages in a stable and efﬁcient way.

BibTex:

@dataset{Yuguang_Yang_and_Johns_Hopkins_University_2024,
    abstract = {Deep reinforcement learning for high dimensional, hierarchical control tasks usually requires the use of complex neural networks as functional approximators, which can lead to inefﬁciency, instability and even divergence in the training process. Here, we introduce stacked deep Q learning (SDQL), a ﬂexible modularized deep reinforcement learning architecture, that can enable ﬁnding of optimal control policy of control tasks consisting of multiple linear stages in a stable and efﬁcient way.},
    author = {Yuguang Yang and Johns Hopkins University},
    doi = {10.57702/y6obe121},
    institution = {No Organization},
    keyword = {'Deep Q learning', 'Hierarchical control', 'Multi-stage control tasks'},
    month = {dec},
    publisher = {TIB},
    title = {A Deep Reinforcement Learning Architecture for Multi-stage Optimal Control},
    url = {https://service.tib.eu/ldmservice/dataset/a-deep-reinforcement-learning-architecture-for-multi-stage-optimal-control},
    year = {2024}
}