Real-world robots are not eusocial ants: Byzantine-resistance in swarm robotics systems using blockchain technology
Speaker: Alexandre Pacheco
Abstract:
A robot swarm is a type of mobile multi-robot system whose design is
inspired by the complex societies of eusocial insects: while an
individual ant cannot accomplish much, colonies composed of millions
have adapted social behaviours that allow them to thrive in all
corners of our planet.
Like their natural counterparts, swarm robotics systems avoid
centralized control architectures, and instead promote local
communications and individual decision making. These design choices
enable a system that is scalable, flexible and fault-resistant by
redundancy. Despite this, recent research has shown that robot swarms
are prone to Byzantine faults (e.g., when communication channels are
tampered, or when faults from one robot can propagate to its peers).
In the absence of supervision or central control, robot swarms that
operate in the real-world may be prone to failures caused by faults,
attacks or exploitation.
Blockchain technology is a new research field that combines concepts
from micro-economics and cryptography to address Byzantine-fault
tolerance in peer-to-peer internet networks. Similarly to the human
agents in blockchain networks, robots can exhibit a wide variety of
behaviors that can be classified as Byzantine. In our research, we
consider that robots may be faulty (due to hardware malfunctions),
malicious (due to tampering from adversaries) or non-altruistic
(because they have been programmed by self-motivated individuals).
In this presentation I will provide a short introduction to swarm
robotics and blockchain technology while diving deeper into the
"oracle problem": how to securely validate the information that is
proposed by potentially Byzantine robots in a decentralized way.
Bio:
Alexandre Pacheco is a Ph.D. student at IRIDIA, the Artificial
Intelligence research laboratory of the Université Libre de Bruxelles,
under the supervision of Professor Marco Dorigo. Before that he
completed his BSc. degree in Mechanical engineering at Instituto
Superior Técnico, in Lisbon, and MSc. degree in Systems and Control
engineering at the Technical University of Delft where he graduated in
2018. Throughout 2017 he was a control systems intern at Intelligent
Systems Division at the NASA Ames Research Center.
Alexandre's PhD research combines swarm intelligence, blockchain
technology and economics, investigating how complex swarm behaviours
emerge from the economic interactions of self-interested robots, and
how a Byzantine-tolerant swarm operation can be achieved using
crypto-economic mechanisms.
Acknowledgment:
This talk is part of our ongoing 2022 bilateral cooperation between
Québec and the Brussels-Capital Region, jointly supported by the
Québec Ministry of International Relations and la Francophonie (MRIF)
and the Brussels Regional Public Service (SPRB).