Hunting algorithm visualization and performance evaluation through BDI agent simulation

This paper presents an agent-based simulation approach to evaluating hunting algorithm performance using Netlogo and the Belief, Desire, and Intent (BDI) architecture. Netlogo was customized to a number of two-dimensional scenarios with both hunting agents and prey agents interacting with each other and with the obstacles within the scenarios. The agents were developed with the flexibility to instantiate many types of hunters and prey. The differences in both the hunting agent and prey agents are with their skills (communications, perception, speed, etc.) and their cognitive abilities. To evaluate the viability of the approach, it was used to evaluate two hunting algorithms: The Lion Optimization Algorithm (LOA) and the Grey Wolf Optimization (GWO) algorithm. The experimental results show that the LOA was more resilient to obstacles than was the GWO. In the presence of obstacles, the lionesses are more reliable in completing joint convergence onto the prey. While the wolves have a lower convergence rate, they display an ability to recover from the confusion caused by obstacles to finally close in on their prey. The research concludes that the NetLogo programming environment is successfully adapted to the BDI architecture and is very effective in structuring a large agent-based system.