A sensitivity analysis of parameters in an agent-based model for crowd simulations

In this research paper, we present a sensitivity analysis of parameters utilized in an agent-based model for crowd simulations. The model is made up of two types of agents that explore a virtual environment to reach an exit from a specified starting point. They behave differently depending on their type, and they may be collaborative, acting carefully to help others reach the exit, or defectors, acting independently and wildly. To simulate the agents' and environment dynamics we used the Ant Colony Optimization algorithm principles. Three metrics to evaluate the effects of these behaviors have been used: the number of exited agents, the path cost, and the exit time. Furthermore, two different types of analyses have been carried out and are presented: group analysis, in which the performance of the groups into which the agents are split are compared; and types' analysis, where the performance of the two types of agents, i.e., collaborators and defectors, are investigated and compared.