Behavior Mining of Robot-Animal Mixed Swarm

Emergent intelligence and collective behaviors have long captivated the interest of researchers. These phenomena often stem from intricate interactions among numerous individuals adhering to straightforward rules. This paper focuses on a mixed swarm consisting of both controllable and uncontrollable swarms. We aim to harness the controllable group's behavior to influence and intervene in the uncontrollable group's actions. Here the controllable group is a robot swarm that can be controlled by human, while the uncontrollable group primarily comprise biological swarms. Designing behavioral rules for the controllable group to effectively affect the behavior of the uncontrolled group presents a challenge. To address this issue, we propose a behavior-mining-based approach. By utilizing the NEAT algorithm through iterative processes, we generate a multitude of potential behavioral rules for the controllable group. Subsequently, we gather data on how the uncontrollable group responds to the actions of the controllable group. We then apply dimensionality reduction and clustering techniques to extract distinctive behavioral trajectory features for the uncontrollable group (method features). Ultimately, through a series of simulation experiments, we validate the effectiveness of this approach.