Predicate Invention for Bilevel Planning

Authors: Tom Silver, Rohan Chitnis, Nishanth Kumar, Willie McClinton, Tomás Lozano-Pérez, Leslie Kaelbling, Joshua B. Tenenbaum

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

Reproducibility Variable Result LLM Response
Research Type Experimental Experimentally, we show across four robotic planning environments that our learned abstractions are able to quickly solve held-out tasks, outperforming six baselines.
Researcher Affiliation Collaboration 1MIT Computer Science and Artificial Intelligence Laboratory 2Meta AI {tslvr, njk, wbm3, tlp, lpk, jbt}@mit.edu, ronuchit@meta.com
Pseudocode Yes Algorithm 1 Bilevel Planning Plan(x0, g, Ψ, Ω, Σ ):
Open Source Code No The paper does not provide an explicit statement or link for open-sourcing the code for the described methodology.
Open Datasets No The paper mentions 'four robotic planning environments' and that 'Demonstrations are collected by bilevel planning with manually defined abstractions' but does not provide concrete access information (link, DOI, specific citation) for a publicly available dataset.
Dataset Splits No The paper mentions 'a set of training tasks' and 'a set of demonstrations D, with one demonstration per task' and '50 evaluation tasks' but does not specify exact split percentages, sample counts for distinct train/validation/test splits, or refer to predefined splits with citations for reproducibility.
Hardware Specification No The paper does not provide specific details on the hardware used, such as exact CPU or GPU models, memory, or cloud instance types.
Software Dependencies No The paper mentions optimizers (Adam) and activation functions (ELU), but does not provide specific software dependencies like library names with version numbers (e.g., PyTorch 1.9, TensorFlow 2.x).
Experiment Setup Yes Here, ϵ > 0 is a small constant (10 5 in our experiments)... a large constant (103 in our experiments)... large constant tupper (105 in our experiments)... wreg = 10 4 in our experiments... ngrammar (200 in experiments).