The PETLON Algorithm to Plan Efficiently for Task-Level-Optimal Navigation

Authors: Shih-Yun Lo, Shiqi Zhang, Peter Stone

JAIR 2020 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments have been conducted both in simulation and on a mobile robot using object delivery tasks in an indoor office environment. The key observation from the results is that PETLON is more efficient than a baseline approach that pre-computes motion costs of all possible navigation actions, while still producing plans that are optimal at the task level.
Researcher Affiliation Collaboration Shih-Yun Lo EMAIL Department of Mechanical Engineering, The University of Texas at Austin; Shiqi Zhang EMAIL Department of Computer Science, The State University of New York at Binghamton; Peter Stone EMAIL Department of Computer Science, The University of Texas at Austin and Sony AI
Pseudocode Yes Algorithm 1 PETLON algorithm Require: sinit, xinit,SG, f, h, Dt, Dm,M, Pt, Pm
Open Source Code No The paper does not provide a direct link to source code, explicitly state code release, or mention code in supplementary materials for the described methodology.
Open Datasets No The paper mentions that the test domain is 'our office environment, with the map pre-scanned and constructed by running the SLAM algorithm (Thrun et al., 2005)', indicating a custom environment without providing access information for a publicly available dataset.
Dataset Splits No The paper describes experimental conditions such as 'different domain scale-up' and 'different numbers of objects' or 'obstacles' for evaluation, but it does not specify any dataset splits like training, validation, or test sets.
Hardware Specification Yes We use a laptop equipped with 2.2GHz i7 processor and 16GB RAM on OS X for all reported results. [...] The robot uses an RMP 110 mobile platform, onboard auxiliary battery, desktop computer (with touchscreen), and Velodyne VLP-16 for perception (Khandelwal et al., 2017).
Software Dependencies No The paper mentions task planners implemented using 'Clingo (Gebser et al., 2014)' and 'Fast Downward (FD) (Helmert, 2006)' but does not provide specific version numbers for these software components.
Experiment Setup Yes For the motion planner, we draw samples with a density of two poses per square meter, and the resultant plan quality has small variations (mostly within one meter) among trials. [...] We manipulated domain visibility by randomly sampling different numbers of obstacles, producing different obstacle densities. Obstacles are of size 0.4m by 0.4m, and are large enough to potentially block the office entrances or prevent the robot from fetching objects in areas.