University of Southampton
Self-organised communication-aware control structures for robot swarms
Abstract
dc:description.abstractRobotic swarms are complex systems that rely on local communication between individual members of the swarm to spread information about their state and the environment. This information informs decisions and aids in tasks such as exploration and mapping. When communications break down between members of a swarm, it can become difficult to maintain accurate and up-to-date information about the state of the swarm and the environment. This problem is pertinent when humans are involved and may act as operators or teammates of the swarm. Here it is vital that the swarm can coordinate to distil and disseminate the vast amounts of information collected to the humans and throughout the swarm effectively, to maintain situational awareness.<br/><br/>The collective decision-making of a swarm is one aspect that relies heavily on the ability to share information and observations to reach a swarm-wide consensus. This thesis investigates how communication constraints affect the swarm’s ability to reach a consensus and implement a communication-aware coordination strategy to mitigate these effects. We propose the communication-constrained collective decision-making problem and compare the performance of several collective decision-making strategies, enhanced with our coordination algorithm. We find that using such an approach improves the speed of a swarm to reach a consensus.<br/><br/>Following on from this work, we examine a hybrid swarm system in a communication-limited environment. While swarms are traditionally considered decentralised systems, recent approaches have integrated decentralised and centralised control into a swarm system. We study the trade-offs in performance and communication in a hybrid system that can vary its control structure. We find that a higher level of centralisation does not guarantee higher performance and study how communication with a human operator is affected by the control structure. This work is extended to assess the feasibility of enabling a swarm system to learn the optimal control structure on the fly, according to mission requirements.
Degree
thesis:*- Name dc:type.qualificationname
- Ph.D.
- Level dc:type.qualificationlevel
- doctoral
- Grantor dc:publisher.institution
- University of Southampton
- Year dc:date.issued
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Kelly, Thomas Graham
- Advisors dc:contributor.advisor
-
- Tarapore, Danesh
- Soorati, Mohammad
- Ramchurn, Gopal
- Zauner, Klaus-Peter
- ORCID person:identifier.orcid
Rights
- Language dc:language
- English, eng