Robustness in Probabilistic Temporal Planning

Authors: Jeb Brooks, Emilia Reed, Alexander Gruver, James Boerkoel

AAAI 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our empirical evaluation demonstrates that our robustness approximations better estimate plan success rate than previous metrics for determining the quality of a schedule.
Researcher Affiliation Academia Harvey Mudd College, Claremont, CA {jbrooks,egreed,agruver,boerkoel}@g.hmc.edu
Pseudocode Yes Algorithm 1: Sampling-based Simulator; Algorithm 2: Representative Simulator
Open Source Code No Control code available upon request.
Open Datasets No The paper describes creating its own empirical distributions ('We empirically derived these pdfs by running our three robots around a one unit square 100 times each...'), but does not provide any access information (link, DOI, specific citation) for this dataset.
Dataset Splits No The paper describes running multi-robot scenarios 50 times but does not specify any training, validation, or test dataset splits in terms of percentages, sample counts, or references to predefined splits.
Hardware Specification Yes We tested our algorithms on an Intel Xeon E5-1603 2.80GHz quadcore processor with 8 GB of RAM running Ubuntu 12.04 LTS.
Software Dependencies No The paper mentions 'Pu LP, a Python library that leverages Coin MP an open source linear programming library' and 'Ubuntu 12.04 LTS', but it does not provide specific version numbers for software dependencies like Python, Pu LP, or Coin MP.
Experiment Setup Yes For the two approximation algorithms, we used the PF and PT pdfs displayed in Figure 5 as our models of durational uncertainty for the PSTN input to both the Sampling-based Simulator (N = 1000) and Representative Simulator. We empirically derived these pdfs by running our three robots around a one unit square 100 times each, and for each independent maneuver, building a histogram using a kernel smoother.