logo SBA


Digital archive of theses discussed at the University of Pisa


Thesis etd-04092011-101435

Thesis type
Tesi di dottorato di ricerca
Thesis title
Swarm Abstractions for Distributed Estimation and Control
Academic discipline
Course of study
tutor Innocenti, Mario
tutor Pollini, Lorenzo
  • consensus algorithms
  • coordination
  • coverage
  • decentralized control
  • decentralized estimation
  • deployment
  • descriptor functions
  • multi-robot
  • robotic networks
  • swarm
  • task-assignment
Graduation session start date
Swarm of vehicles, instead of single or small groups of vehicles, are becoming
of large interest in a variety of applications for their intrinsic robustness and flexibility.
Probably, the most difficult issue is coordination and control of a large
number of small vehicles with limited processing, communication and power capabilities.
A large number of solutions for practical problems have been studied and proposed
by researchers worldwide; they are often specific solutions to specific problems
that are difficult to generalize. Most of them cannot manage the natural
heterogeneity of a large group of vehicles: it is often desired that
a swarm performs a complex mission in different phases
that may require different specialized vehicles instead of a multi role one.

The aim of this thesis is to develop a framework for swarm modeling, decentralized estimation
and control capable of managing many of the the application fields
proposed and studied in the literature. The environment and mission(s) under consideration
are as general as possible, as well as the characteristics of the agents
which may be all identical or heterogeneous.

The Descriptor Functions Framework, the main topic and contribution of this thesis,
associates a specific function to each agent, and one or more functions to the mission and uses
an analytical framework for generation of the agents control law.
Decentralized control and estimation of the relevant variables needed
as feedback are cast into an optimization problem; analysis of equilibria
and formal proof of convergence of the estimation and control law proposed are derived.
Then, a bio-inspired method for task self-assignment is applied to the Framework;
swarm vehicles are given the capability to select the best task to perform,
swap task with other agents during the mission, and assess the degree of completeness
of the various tasks in order to balance the swarm capabilities among them.
Finally an hardware test bed, together with experimental results, is be presented.