Plan Recognition in Continuous Domains

Authors: Gal Kaminka, Mor Vered, Noa Agmon

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

Reproducibility Variable Result LLM Response
Research Type Experimental We provide formal arguments for the usefulness of mirroring, and empirically evaluate mirroring in more than a thousand recognition problems in three continuous domains and six classical planning domains.
Researcher Affiliation Academia Gal A. Kaminka, Mor Vered, Noa Agmon Computer Science Department Bar Ilan University, Israel {galk,veredm,agmon}@cs.biu.ac.il
Pseudocode No The paper does not contain any pseudocode or algorithm blocks.
Open Source Code No The paper does not provide concrete access to source code for the methodology described. It mentions using open-source libraries and simulators like OMPL, ROS, and Gazebo, but not its own implementation code.
Open Datasets Yes We used the Open Motion Planning Library (OMPL (S ucan, Moll, and Kavraki 2012)) cubicles environment along with the default robot... We used the entire set of benchmark plan-recognition problems used in (Ram ırez and Geffner 2010) and then in (Sohrabi, Riabov, and Udrea 2016; Pereira, Oren, and Meneguzzi 2016)... We additionally evaluated our continuous goal recognizer on the shape sketch recognition domain introduced in (Vered, Kaminka, and Biham 2016).
Dataset Splits No The paper does not explicitly state training, validation, or test dataset splits. It describes generating recognition problems or using existing benchmarks, but not the specific partitioning of data into these subsets.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, memory) used for running its experiments. It mentions a 'robot' and 'simulator' but no hardware specifications.
Software Dependencies No The paper mentions software like OMPL, ROS, Gazebo, and the hspf planner, but it does not specify version numbers for these or any other ancillary software components.
Experiment Setup Yes Each call to the planner was given a time limit of 1 sec. For cost, we measured the length of the path. We set 11 points spread through the cubicles environments. We then generated two observed paths from each point to all others, for a total of 220 = 110 * 2 recognition problems. Each plan consisting of between 20-75 points. The observations were obtained by running the asymptotically optimal planner RRT* on each pair of points (time limit of 5 minutes).