In this talk, we will first discuss some of the characteristic features of swarm robotics, a new, promising sub-area of distributed robotics dealing with large numbers of simple robotic nodes. We will then illustrate with concrete examples the challenges posed by the design and the control of these distributed, often substantially stochastic systems, and the limitations of more conventional model-based approaches for tackling such challenges. As potential answer to such limitations, we will present methods based on multiple levels of abstraction we have developed to evaluate, control, and optimize the performances of swarm robotic systems. One important feature of these methods is their ability to deal with the broad variety of length and time scales as well as technological substrates encountered in this type of systems. We will support the discussion with a recent case study which originated from the seminal stick-pulling experiment. This case study is characterized by a more flexible and generalizable set-up and is better suited to the study of swarms of miniature robots (2 cm in length) endowed with limited sensing, computing, and communication capabilities. We will conclude the talk by illustrating how such multi-level modeling approach can be leveraged to go beyond swarm robotic systems and be applied to distributed (intelligent) micro-systems.
Grégory Mermoud and Alcherio Martinoli
Distributed Intelligent Systems and Algorithms Laboratory
EPFL ENAC IIE DISAL
E-mail: gregory.mermoud_at_epfl.ch and alcherio.martinoli_at_epfl.ch