Consensus-based Control Over Wireless Channels
vorgelegt von
M. Sc.
Fabio Molinari
ORCID: 0000-0003-2617-962X
an der Fakultät IV - Elektrotechnik und Informatik
der Technischen Universität Berlin
zur Erlangung des akademischen Grades
Doktor der Ingenieurwissenschaften
-Dr.-Ing.-
genehmigte Dissertation
Promotionsausschuss:
Vorsitzender: Prof. Dr.-Ing. Marc Toussaint
Gutachter: Prof. Dr.-Ing. Jörg Raisch
Gutachter: Prof. Riccardo Scattolini
Gutachter: Prof. Dr.-Ing. Sławomir Stańczak
Tag der wissenschaftlichen Aussprache: 15. Dezember 2021
Berlin 2022
In memory of Mariuccia and Franco.
Abstract
Consensusisausefulconceptinsituationswhere agroupofautonomous
agents seek an agreement over a variable of common interest for the
whole system. Agents communicate and, based on the information
they collect from others, they take individual decisions according to
the specific consensus protocol they implement. A consensus proto-
col that needs less resources for achieving a given result is more ef-
ficient. Efficiency in consensus has become a hot topic in research
during the last years, mainly due to the rising number of applications
that require battery-powered agents. This thesis proposes to increase
the efficiency of consensus by exploiting the interference property of
the wireless channel. This phenomenon has always been combatted.
In fact, signals simultaneously transmitted by multiple agents in the
same frequency are attenuated by unknown coefficients and summed
up at the receiver. The standard approach is to discard such corrupted
received signal, although it carries some useful information. In this
thesis, we show that this corrupted signal can be actually used for
achieving consensus. Under the assumption of a real-valued fading
channel, we propose average consensus and max-consensus protocols
that exploit interference. They both exhibit better performance than
standard methods in terms of required wireless resources. We show
that such protocols can be used for two practical problems, i.e., forma-
tion control of nonholonomic robots and distributively solving linear
algebraic equations. This confirms the practical benefits of exploiting
interference and motivates future experimental implementation.
Loading more pages...