Deceptive Path-Planning

Authors: Peta Masters, Sebastian Sardina

IJCAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, we provide an empirical evaluation, review related work and present our conclusion. and Figure 4: Results show the percentage of paths returned by each strategy that were deceptive when tested at 10%, 25%, etc., of their path length prior to the RMP (beyond the RMP, all paths are truthful). Table columns show average (total) path costs and average time taken to generate the (total) path.
Researcher Affiliation Academia Peta Masters and Sebastian Sardina RMIT University, Melbourne, Australia {peta.masters, sebastian.sardina}@rmit.edu.au
Pseudocode No The paper discusses formulas and strategies but does not contain structured pseudocode or algorithm blocks.
Open Source Code No The paper does not provide any concrete access to source code for the methodology described.
Open Datasets Yes We generated a problem set based on game maps from the Moving-AI benchmarks [Sturtevant, 2012]
Dataset Splits No The paper states 'For each of 50 problems, we generated one optimal path using a standard implementation of A* and four deceptive paths (each using a different strategy),' but it does not specify any explicit training, validation, or test dataset splits.
Hardware Specification Yes Experiments were conducted on a i7 3.6GHz machine with 8GB RAM.
Software Dependencies No The paper mentions 'a standard implementation of A*' and 'Ramirez and Geffner s method of goal recognition,' but does not list any specific software dependencies with version numbers.
Experiment Setup Yes We generated a problem set based on game maps from the Moving-AI benchmarks [Sturtevant, 2012] to which we added three extra candidate goals at random locations. For each of 50 problems, we generated one optimal path using a standard implementation of A* and four deceptive paths (each using a different strategy). We timed path generation and recorded path costs. We truncated paths at the RMP (beyond which all paths would be truthful) and, using Ramirez and Geffner s method of goal recognition, calculated probabilities at intervals to confirm/assess deceptive density and extent. and if h(n, gr) < h(n, gmin) then h(n, t) = αh(n, t), where constant α > 1.