Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
Autonomous Target Search with Multiple Coordinated UAVs
Authors: Chiara Piacentini, Sara Bernardini, J. Christopher Beck
JAIR 2019 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Our extensive experimental evaluation studies the scalability of the different techniques and identifies the conditions under which one approach becomes preferable to the others. |
| Researcher Affiliation | Academia | Chiara Piacentini EMAIL Department of Mechanical and Industrial Engineering University of Toronto, Toronto, Canada, ON M5S Sara Bernardini EMAIL Department of Computer Science Royal Holloway University of London Egham, Surrey, UK, TW20 0EX J. Christopher Beck EMAIL Department of Mechanical and Industrial Engineering University of Toronto, Toronto, Canada, ON M5S |
| Pseudocode | Yes | The greedy procedure, outlined in Algorithm 1, proceeds iteratively to construct a set of patterns according to the recursive structure of Equation (1). The assignment algorithm (Algorithm 2) runs through all the possible observers and, for each observer (Algorithm 2: line 2), iterates over the sequence of the search patterns that are already assigned to it in the time window associated with the new pattern (Algorithm 2: line 3). ... Starting from the partial plan π necessary to achieve a state, Algorithm 3 calculates the remaining part of the plan ˆπ, based on the sequence of search patterns that the greedy algorithm would produce (Algorithm 1) and returns the heuristic value of the state and the number of actions in ˆπ. ... Instead of the greedy algorithm presented in Section 4.1, which is quadratic in the number of search pattern candidates, we consider an accelerated (linear) version of it, following the procedure presented by Minoux (1978) (Algorithm 4). |
| Open Source Code | No | The paper discusses various algorithms and models, including extensions to PDDL and the use of specific solvers like popf-tif and IBM ILOG CPLEX CP Optimizer. However, there is no explicit statement or link provided by the authors indicating that their *own* implementation code for the methodologies described in this paper is publicly available. |
| Open Datasets | No | We generate the initial set of patterns C by running a Monte Carlo simulation (MCS) over the area of operation. ... We model the target motion with a standard MCS. ... The target follows a route acquired using Graph Hopper (Karich, 2015). ... The paper describes how the authors generate their search patterns and target routes using simulations and a tool (Graph Hopper), but does not provide access information (link, citation, repository) to any specific dataset they used or generated for public access. |
| Dataset Splits | No | For the first experiment, we randomly generate 40 instances with 30 candidate search patterns. For each instance, we reduce the number of search patterns to a minimum of six, and we vary the number of observers, from a minimum of one to a maximum of five. ... For every configuration, we simulate 500 missions. ... The paper describes how experimental instances were generated (e.g., 40 random instances, 500 simulated missions) and parameters varied (e.g., number of search patterns, number of observers), but it does not specify traditional dataset splits like training, validation, or test sets in the context of machine learning or statistical evaluation. |
| Hardware Specification | Yes | We run all the experiments on a Xeon 3.5GHz processor machine running Mac OS X Sierra. |
| Software Dependencies | Yes | We solve the CP model using IBM ILOG CPLEX CP Optimizer v12.8.0 (Laborie et al., 2018). ... We use the planner popf-tif (Piacentini, Alimisis, Fox, & Long, 2015) to build plans for the UAVs. popf-tif is based on the partial order temporal planner popf2 (Coles et al., 2010) ... The PDDL (Fox & Long, 2003) model that we use presents a set of objects... The target follows a route acquired using Graph Hopper (Karich, 2015). |
| Experiment Setup | Yes | All solvers are given one minute to generate the plans. ... We use a time-discretization of 10 and 100 seconds, and we indicate these two variations as CP 10 and CP 100, respectively. ... We call CPws iv the model that is warm-started with the G solution. ... For every configuration, we simulate 500 missions. ... The search area considered in the MCS is an angular sector from the LKP. When only one observer is available the angle is 120 and it increases uniformly as more observers are available, up to a maximum of 180 for 5 observers. The MCS considers 20 time points and, for each time point, it generates up to |O| + 1 search pattern candidates. |