Density Constrained Reinforcement Learning

Authors: Zengyi Qin, Yuxiao Chen, Chuchu Fan

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We use a set of comprehensive experiments to demonstrate the advantages of our approach over state-of-the-art CRL methods, with a wide range of density constrained tasks as well as standard CRL benchmarks such as Safety-Gym.Comprehensive experiments are conducted on various density constrained problems, which demonstrate the advantages of our approach over state-of-the-art CRL methods, in terms of satisfying system specifications and improving the cumulative reward.
Researcher Affiliation Academia 1Massachusetts Institute of Technology 2California Institute of Technology.
Pseudocode Yes Algorithm 1 Template of the DCRL algorithm
Open Source Code No The paper does not provide a public link or explicit statement about the release of its own source code.
Open Datasets Yes The standard CRL benchmarks are from Mu Jo Co and Safety-Gym (Ray et al., 2019). It is adopted from Blahoudek et al. (2020) and is shown in Figure 1.
Dataset Splits No The paper does not provide specific details on train/validation/test dataset splits, percentages, or explicit references to predefined splits for reproducibility.
Hardware Specification No The paper does not provide specific hardware details (e.g., exact GPU/CPU models, memory amounts) used for running its experiments.
Software Dependencies No The paper mentions software components like 'TRPO', 'DDPG', 'CPO', 'PCPO', 'RCPO', 'Mu Jo Co', and 'Safety-Gym' but does not specify their version numbers.
Experiment Setup No The paper states that 'All the three baseline approaches and our DCRL have the same number of neural network parameters' but does not provide specific hyperparameters such as learning rates, batch sizes, or detailed optimizer settings for reproducibility.