Autonomous hierarchical multi-level clustering for multi-uav systems

Jonathan Ponniah, Mirco Theile, Or D. Dantsker, Marco Caccamo

Research output: Chapter in Book/Report/Conference proceedingConference contribution


The strategy of clustering is introduced to enable coordination in decentralized multi-UAV systems. Each cluster is an organized unit comprised of several cluster-members and one cluster-head. We propose the concept of multi-level clustering, in which cluster-heads form successively higher-level clusters, resulting in a tree-shaped hierarchy. Multi-level clustering provides a mechanism for aggregating local states and disseminating the information needed for system coordination. Related work shows that aggregate information is beneficial for efficient UAV path planning using reinforcement learning. We propose rules for scalable multi-level cluster-formation, taking into consideration the computational and communication loads associated with cluster maintenance and information aggregation and dissemination. The viability of the proposed concept is demonstrated in preliminary simulations. The scenarios considered examine the effects of agent motion, takeoff, and landing on multi-level clustering. The simulation results show that multi-level clustering is robust to the dynamics of multi-UAV environments.

Original languageEnglish (US)
Title of host publicationAIAA Scitech 2021 Forum
PublisherAmerican Institute of Aeronautics and Astronautics Inc, AIAA
Number of pages12
ISBN (Print)9781624106095
StatePublished - 2021
Externally publishedYes
EventAIAA Science and Technology Forum and Exposition, AIAA SciTech Forum 2021 - Virtual, Online
Duration: Jan 11 2021Jan 15 2021

Publication series

NameAIAA Scitech 2021 Forum


ConferenceAIAA Science and Technology Forum and Exposition, AIAA SciTech Forum 2021
CityVirtual, Online

ASJC Scopus subject areas

  • Aerospace Engineering


Dive into the research topics of 'Autonomous hierarchical multi-level clustering for multi-uav systems'. Together they form a unique fingerprint.

Cite this