TACO: Learning Task Decomposition via Temporal Alignment for Control

Authors: Kyriacos Shiarlis, Markus Wulfmeier, Sasha Salter, Shimon Whiteson, Ingmar Posner

ICML 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We evaluate the approach on multiple domains, including a simulated 3D robot arm control task using purely image-based observations. The approach performs commensurately with fully supervised approaches, while requiring significantly less annotation effort, and significantly outperforms methods which separate segmentation and imitation.
Researcher Affiliation Academia 1Informatics Institute, University of Amsterdam, Netherlands 2Department of Engineering Science, University of Oxford, United Kingdom 3Department of Computer Science, University of Oxford, United Kingdom.
Pseudocode No The paper describes algorithms and derivations but does not include any structured pseudocode or algorithm blocks.
Open Source Code No The TACO algorithm and experiments will soon be available at https://sites.google.com/view/taco-ml.
Open Datasets No The paper describes the Nav-World, Craft, and Dial domains used in experiments, but does not provide specific links, DOIs, or formal citations for public access to these datasets. For Nav-World and Dial, they appear to be custom-simulated environments. For Craft, while it mentions 'Andreas et al. (2017) introduced the Craft Domain', it doesn't provide access to the dataset itself.
Dataset Splits No The paper specifies training dataset sizes (e.g., '50, 400, and 1000 demonstrations') and describes the test setting ('unseen, longer tasks of length Ltest = 4 (zero-shot setting)'), but it does not explicitly mention a separate 'validation' dataset split or its size/percentage.
Hardware Specification No The paper mentions running experiments in a 'simulated 3D robot arm control task' and a 'JACO 6 Do F manipulator simulated in Mu Jo Co', but it does not provide any specific details about the hardware (e.g., GPU/CPU models, memory) used for these simulations or training.
Software Dependencies No The paper mentions 'Mu Jo Co (Todorov et al., 2012)' as a simulation engine, but it does not specify a version number for Mu Jo Co or any other software dependencies with version numbers.
Experiment Setup No The paper states 'Details of the architectures used can be found in the Appendix.' which implies specific experimental setup details exist, but these are not provided in the main text of the paper.