When to Sense and Control? A Time-adaptive Approach for Continuous-Time RL

Authors: Lenart Treven, Bhavya , Yarden As, Florian Dorfler, Andreas Krause

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate that state-of-the-art RL algorithms trained on TACOS drastically reduce the interaction amount over their discrete-time counterpart while retaining the same or improved performance, and exhibiting robustness over discretization frequency.
Researcher Affiliation Academia Lenart Treven , Bhavya Sukhija, Yarden As, Florian Dörfler, Andreas Krause ETH Zurich, Switzerland
Pseudocode No The paper provides mathematical formulations and definitions but does not include a clearly labeled pseudocode or algorithm block.
Open Source Code Yes We provide our implementation at https://github.com/lasgroup/TaCoS.
Open Datasets Yes We consider the RC car (Kabzan et al., 2020), Greenhouse (Tap, 2000), Pendulum, Reacher, Halfcheetah and Humanoid environments from Brax (Freeman et al., 2021).
Dataset Splits No The paper does not explicitly state the training, validation, or test dataset splits with specific percentages, counts, or methodology for reproduction.
Hardware Specification No The paper does not provide specific details regarding the hardware used for running experiments (e.g., GPU/CPU models, memory, or cloud provider specifications).
Software Dependencies No The paper mentions the use of algorithms like SAC and PPO but does not provide specific version numbers for any software dependencies, libraries, or frameworks used.
Experiment Setup Yes We investigate both the bounded number of interactions and interaction cost settings in our experiments. In particular, we study how the bound K affects the performance of TACOS and compare it to the standard equidistant baseline. We further study the interplay between the stochasticity of the environments (magnitude of g ) and interaction costs and the influence of tminon TACOS. For all experiments in this section, we combine SAC with TACOS (SAC-TACOS).