Comparison of Desynchronization Methods for a Decentralized Swarm on a Logistical Resupply Problem

Experiments are divided into response threshold and response duration segments, each with a separate ANOVA for each benchmark. 7 probability distributions are used to generate threshold values: constant, uniform, Gaussian (μ = 0.50, 0.25), and Poisson (λ = 3, 5, 7). For response duration, we test prob_check values from [0.1, 1.0] in increments of 0.1. Inputs contain 10 runs of distinct schedules. Measure stress_index is defined as agent resources relative to task demands, being directly proportional to Scaling_factor , and inversely proportional to Popsize. • 8 experiments [3] [A,B,C] = Decreasing stress_index via decreased Scaling_factor [D,A,E,F,G] = Decreasing stress_index via increased Popsize All experiments are run on long and short schedules, for response duration, we include ushort, as well. The Two-Way ANOVA examines 1) Main effects: • Threshold/Duration Level • Schedule Type 2) Interaction effects between both factors