Analyzing the Parameters of Prey-Predator Models for Simulation Games

Key insights

6 takeaways

01
Carrying capacity in the base configuration is about 158, and it rises almost linearly with food growth rate before levelling off at a limit set by the food-growth-delay parameter.See it on page 5
02
The four LV parameters for the base configuration were estimated as alpha = 0.075, beta = 0.00543, gamma = 0.0689, and delta = 0.0031.See it on page 7
03
The authors propose beta as a competitive fitness metric, since a growth in beta indicates improving predator hunting skills and a decrease indicates prey better evading predation.See it on page 5
04
For the more intense configuration, the predicted first predator peak is 42, while the observed predator peak is 46.See it on page 7
05
The authors state that the predictions are in reasonable agreement with the observations, accounting for the stochasticity of the simulation.See it on page 7
06
The authors suggest that prey may fare better roaming in smaller herds that split and explore different directions, since large herds could exhaust grazing areas quickly.See it on page 8

Summary

This paper, by Seongdong Kim, Christoph Hoffmann and Varun Ramachandran from Kaywon School of Art & Design and Purdue University, describes emergent behavior in a class of prey-predator simulation models used in simulation games. The simulation uses rule-based agent behavior built on the OpenSteer library and is modulated by user-assigned parameters. The authors relate these parameters to the extended Lotka-Volterra (LV) equations, modified to include a carrying capacity for prey, and estimate four key parameters (alpha, beta, gamma, delta) from simulation runs using a method attributed to Sharov.

In the base configuration, the carrying capacity is about 158, and the estimated parameters are alpha = 0.075, beta = 0.00543, gamma = 0.0689, and delta = 0.0031. The authors report that carrying capacity rises almost linearly with food growth rate at first, then approaches a limit set by the food-growth-delay parameter. A more intense world with alpha = 0.21, beta = 0.024, gamma = 0.23, and delta = 0.002 was integrated and compared with observations.

For that intense configuration, the predicted first predator peak is 42 after about 40 time steps, while the observed predator peak is 46. The authors describe the predictions as in reasonable agreement with observations, allowing for stochasticity and integer values in the simulation.