Detection of Plan Deviation in Multi-Agent Systems

Authors: Bikramjit Banerjee, Steven Loscalzo, Daniel Thompson

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental evaluation in two domains one existing and another new not only establishes the relative tractability of CBEM, but also the validity of its inferences. We evaluate CBEM in two domains: multi-agent tiger (Nair et al. 2003), and drone search a new domain that we introduce here. Figure 4: Left: Per history (θT i ) runtime of CBEM and EM on Multi-agent Tiger for horizons T = 3, . . . , 7.
Researcher Affiliation Collaboration Bikramjit Banerjee School of Computing University of Southern Mississippi Hattiesburg, MS 39402 t Bikramjit.Banerjee@usm.edu Steven Loscalzo AFRL Information Directorate 26 Electronic Parkway Rome, NY 13441 Steven.Loscalzo@us.af.mil Daniel Lucas Thompson School of Computing University of Southern Mississippi Hattiesburg, MS 39402 Daniel.L.Thompson@eagles.usm.edu
Pseudocode No The paper provides mathematical equations and descriptions of the algorithms (EM and CBEM) but does not include a distinct block labeled as 'Pseudocode' or 'Algorithm'.
Open Source Code No The paper does not provide any information or links regarding the availability of its source code.
Open Datasets Yes We evaluate CBEM in two domains: multi-agent tiger (Nair et al. 2003), and drone search a new domain that we introduce here.
Dataset Splits No The paper discusses evaluating the approach in specific domains (multi-agent tiger, drone search) and over different time horizons, but it does not provide specific details on training, validation, or test dataset splits.
Hardware Specification No The paper mentions generally 'This is well beyond the capability of any personal computer existing today' in relation to computational complexity, but it does not specify any particular hardware (GPU, CPU models, etc.) used for running its experiments.
Software Dependencies No The paper does not provide specific details about ancillary software, such as library names or versions, that would be needed to replicate the experiments.
Experiment Setup Yes For each scenario, we use initial δ0 k = 0.9, k = 0, . . . , 9. Figure 3 shows a part of the manually designed layered controller used for Drone 1. This domain has |S| = 256 joint drone locations. Each drone can move in one of the 4 cardinal directions, or send a communication, or stop, for a total of |Ai/j| = 6 actions. The number of observations is |Ωi/j| = 4... Furthermore, T = 10...