Emergence of collective adaptive response based on visual variation

Jingtao Qi, Liang Bai, Yingmei Wei, Huaxi Zhang, Yandong Xiao

Abstract: Models are constantly being developed to replicate the collective behavior of biological groups. However, with popular vision-based models, it is difficult to capture the inter-individual differences in velocity from visual information; consequently, a vision-based group cannot produce an adaptive response to the movements of individuals within the group and reach consensus. To remedy this deficiency, based on the OODA (observation-orientation-decision-action) framework as an abstraction of collective behavior, this paper introduces an alignment response with visual variation as input to a vision-based model and proposes a method based on an adaptive mechanism for generating a collective adaptive response. Through experiments on coordinated collective behavior, the introduction of the proposed alignment response is demonstrated to shorten the time required for the vision-based group to reach consensus, and the polarization increases by 59.42%. In addition, cases of a group guided by an informed individual either performing a simple turn or following one of two complex trajectories are evaluated to verify the generation of a collective adaptive response. The introduction of the adaptive mechanism results in an 11.92% increase in the response. This study remedies the deficiencies of vision-based models and offers a theoretical foundation for addressing communication interference in unmanned swarms.