Design of Influencing Agents for Flocking in Low-Density Settings

Abstract

Flocking is a coordinated collective behavior that results from local sensing between individual agents who have a tendency to orient towards each other. Flocking is common amongst animal groups and could also be useful in robotic swarms. In the interest of learning how to control flocking behavior, several pieces of recent work in the multiagent systems literature have explored the use of influencing agents for guiding flocking agents to face a target direction. However, the existing work in this domain has focused on simulation settings of small areas with toroidal shapes. In such settings, agent density is high, so interactions are common, and flock formation occurs easily. In our work, we study new environments with lower agent density, wherein interactions are more rare. We study the efficacy of placement strategies and influencing agent behaviors drawn from the literature, and find that the behaviors that have been shown to work well in high-density conditions tend to be much less effective in the environments we introduce. The source of this ineffectiveness is that the influencing agents explored in prior work tended to face directions optimized for maximal influence, but which actually separate the influencing agents from the flock. We find that in low-density conditions maintaining a connection to the flock is more important than rushing to orient towards the target direction. We use these insights to propose new influencing agent behaviors that overcome the difficulties posed by our new environments. The best influencing agents we identify act like normal members of the flock to achieve positions that allow for control, and then exert their influence. We dub this strategy ``follow-then-influence.'' We also tackle this problem by using genetic programming to evolve ad hoc behaviors for influencing agents. We use a hand-constructed domain-specific language and evolve populations in a small test environment, before testing the best candidates in larger scenarios. We find that the best genetic behaviors can do as well as our best hand-designed algorithms, since they can strike a balance between quickly turning their neighbors towards the goal direction while not losing influence by flying away from their neighbors.