Efficient Exploration in Resource-Restricted Reinforcement Learning

Authors: Zhihai Wang, Taoxing Pan, Qi Zhou, Jie Wang

AAAI 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments demonstrate that the proposed RAEB significantly outperforms state-of-the-art exploration strategies in resource-restricted reinforcement learning environments, improving the sample efficiency by up to an order of magnitude.
Researcher Affiliation Academia 1CAS Key Laboratory of Technology in GIPAS, University of Science and Technology of China 2Institute of Artificial Intelligence, Hefei Comprehensive National Science Center
Pseudocode No Due to limited space, we summarize the procedure of RAEB in Appendix B.
Open Source Code No The paper does not provide any explicit statements about releasing source code or links to a code repository.
Open Datasets Yes To compare RAEB with the baselines, we design a range of robotic delivery and autonomous electric robot tasks based on Gym (Brockman et al. 2016) and Mujoco (Todorov, Erez, and Tassa 2012).
Dataset Splits No The paper mentions evaluating policies every 10000 training steps but does not specify details about dataset splits for training, validation, or testing.
Hardware Specification No The paper does not specify any hardware details (e.g., GPU/CPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions Gym and Mujoco as platforms but does not provide specific version numbers for these or any other software dependencies.
Experiment Setup Yes For all environments, we use the intrinsic reward coefficient β = 0.25. We use α = 0.25Imax for delivery tasks, α = 2.5Imax for tasks with limited electricity, and α = [0.25Imax, 2.5Imax] for delivery tasks with limited electricity.