The influence of limited visual sensing on the Reynolds flocking algorithm
The interest in multi-drone systems flourished in the last decade and their application is promising in many fields. We believe that in order to make drone swarms flying smoothly and reliably in real-world scenarios we need a first intermediate step which consists in the analysis of the effects of limited sensing on the behavior of the swarm. In nature, the central sensor modality often used for achieving flocking is vision. In this work, we study how the reduction in the field of view and the orientation of the visual sensors affect the performance of the Reynolds flocking algorithm used to control the swarm. To quantify the impact of limited visual sensing, we introduce different metrics such as (i) order, (ii) safety, (iii) union and (iv) connectivity. As Nature suggests, our results confirm that lateral vision is essential for coordinating the movements of the individuals. Moreover, the analysis we provide will simplify the tuning of the Reynolds flocking algorithm which is crucial for real-world deployment and, especially for aerial swarms, it depends on the envisioned application. We achieve the results presented in this paper through extensive Monte-Carlo simulations and integrate them with the use of genetic algorithm optimization.
Screenshot 2022-09-26 at 16.29.42.png
Thumbnail
openaccess
copyright
30.93 KB
PNG
d559041c42ea5720d339a077a8107db3
soria_influence_2018.pdf
openaccess
4.28 MB
Adobe PDF
925b52a8311082fd9101350d6a13c97c