This thesis extends the agent-based HoPE simulator to study the collective evasion of pigeon flocks under predation pressure. It develops six predator attack strategies and two prey evasion rules and incorporates realistic perception factors including visual occlusion and heterogeneity among individuals within the flock. The model provides parametric control of flock size, attack geometry, and agent movement and perception parameters. Simulations are analyzed using the metrics of topological stability Qm(t) and spatial diffusion R^2(t) together with performance indicators such as time to first capture, capture rate, and number of target switches. Systematic experiments across flock sizes N = 10, 30, 100 and different approach angles reveal clear patterns. Occlusion reliably increases the frequency of target switches and significantly prolongs the time to the first capture. Group heterogeneity does not raise the predator’s probability of tracking errors, yet it likewise lengthens the time to capture. The centroid-seeking attack most strongly reduces the flock’s topological stability. This work contributes an extended simulation framework and a reproducible analysis pipeline for the quantitative comparison of strategies.
|
