IRRATIONAL DECISIONS IN THE SWARMS OF AUTONOMOUS AGENTS

The report addresses the behaviour of the mobile agents acting in swarm. The agents are moving toward the same target point known to all the agents. The goal of the swarm is to minimize the energy (or, that is the same, the time or the overall number of steps), by which at least one of the agents reaches the target point.

It is clear that if each agent makes rational decisions and always minimizes its individual energy, then there exist such configurations of the agents’ positions that no one of them reaches the target. That leads to a certain level of randomness in the agents’ motion. However, the direct introduction of random decisions does not always result in reaching the target using minimal energy.

Fortunately, recent studies of the swarms of living organisms demonstrate that in addition to the randomness the agents demonstrate irrational altruism in such a sense that in some arbitrary moments they avoid moving toward the target and allow the other agents to make their steps. We implement these observations in navigation of the swarm of mobile robots.

In the report, we present the generic model of irrational random activity of the agents in swarm and demonstrate that certain level of randomness and irrationality in the individual activity of the agents improves the overall swarm performance; in particular, it decreases the overall energy required to reach the target point by at least one swarm member.