Revisiting Parameters of Bioinspired Behavior Models in Group Foraging Modeling

摘要

Using bioinspired models and methods is one of approaches for solving tasks of swarm robotics. In this paper one of such tasks, modeling of foraging, and it’s solving by creating analogues of social structures of ants and models of feeding behavior are considered. The most important characteristics of ants’ colonies for modeling were defined – individuals number in society and it’s structure, workers’ speed, a communication distance and working area size. Besides, existing experimental basis (a group of robots and a polygon) was estimated for a usage as a hardware platform for experiments. Several models of feeding behavior were considered: a model without foragers’ functions differentiation and a model with differentiation on active and passive ones. Active foragers look for resources by themselves, then they involve passive foragers; passive foragers are settled on a base, while are not involved in harvesting. A set of finite state machines describe the behavior of agents: basic automatons (provide basic behavior functions) and a meta- automaton, that switches with some conditions an execution of basic automatons. Basic movements were tested on experimental basis. A complex test of models were conducted in a simulation program Kvorum. An analogue of real polygon was made in the program. Modeling consists of series of experiments for every model in which agents must harvest resources. Series differ from each other by number of agents. For models’ quality estimation a ratio of received energy to average obtaining time. Experiments settle that model with functions differentiation works more effective.