(Two-day workshop – Building intelligent swarms of drones)

November 13-14, 2017
Manufacture des Tabacs – Room MS 001, S Building
(Access map & Map of the Manufacture available at the bottom of the page)

Université Toulouse Capitole
21 Allée de Brienne,
31000 Toulouse, France

Aim and Scope:

Designing swarms of autonomous mini or micro-drones able to self-organize, sense their environment, coordinate their movements and cooperate to perform collective tasks in real-world situations such as monitoring forest fires and rescuing in emergency situations (e.g.: flooding or earthquake), is a major challenge in collective robotics. This problem has been tackled from different perspectives and by many research fields that often follow different strategies and objectives. In quantitative and computational ethology, recent advances in the study of collective motion and information processing in animal groups such as swarms of insects, schools of fish or flocks of birds has offered new sources of inspiration to design distributed control algorithms for swarms of drones. In these animal groups simple social interactions facilitate the transfer of information between individuals and their ability to quickly respond to changes in the environment. Could these interactions be embodied in artificial swarms and be used to generate adaptive collective behaviors? In computer science, a major challenge resides in the provision of robust and truly decentralized control of the swarm at different levels in the architecture. Micro-drones are unreliable at individual scale, due to low resources and harsh running conditions. Yet, the inherent redundancy of the swarm is an opportunity to provide resilient swarm services. In collective robotics, a crucial issue is the relative perception problem when it comes to implementing flying swarms in real life. Handling a swarm with a global infrastructure (GPS, global communication) is not a suitable solution when envisioning swarms of hundreds or thousands of drones. Scalability of swarms can only be attained through the use of robot-to-robot perception that has to be both reliable and fast to allow as agile maneuvers as is attained by flocks of birds or schools of fish.

This workshop is intended to promote an interdisciplinary approach of distributed control and adaptive collective behavior in swarms of drones and to favor cross-disciplinary interactions between various communities: quantitative ethology, computer science, information technologies, collective and swarm robotics and statistical physics. In particular we intend to discuss the following topics:

1) Control of swarms: adaptation, decentralized decision and local implementation;

2) Mastering uncertainty in swarms: local/global perception, pattern formation and evolution;

3) Multi-level robustness of swarms: robust local control, fault-tolerant coordination and consistency of global planning.

The worshop will take place at the Université Toulouse 1 Capitole and is supported by The Fondation de Coopération Scientifique Sciences et Technologies pour l’Aéronautique et l’Espace (http://www.fondation-stae.net/), by the Toulouse Institute for Complex Systems Studies (https://xsys.fr/), by the Computer Science Research Institute of Toulouse (https://www.irit.fr) and by the Research Centre on Animal Cognition (http://cognition.ups-tlse.fr).


  • Frédéric Amblard – Toulouse Institute of Computer Science Research, CNRS & University Toulouse 1 Capitole
  • Bertrand Jouve – FRAMESPA & Institut de Mathématiques de Toulouse, CNRS, University of Toulouse & XSYS
  • Céline Parzani – Ecole Nationale de l’Aviation Civile
  • Matthieu Roy – Laboratoire d’Analyse et d’Architecture des Systèmes, CNRS
  • Stéphane Sanchez – Institut de Recherches en Informatique de Toulouse, CNRS & Université Toulouse 1 Capitole
  • Guy Theraulaz – Centre de Recherches sur la Cognition Animale, CNRS, Université Paul Sabatier Toulouse

& Support : Aurélie Cretté – XSYS

Speakers :

  • Spring Berman, Autonomous Collective Systems Laboratory, Arizona State University, Tempe, USA
  • Daniel Delahaye, Ecole Nationale de l’Aviation Civile, Toulouse, France
  • Marco Dorigo, IRIDIA, Université Libre de Bruxelles, Brussels, Belgium
  • Eliseo Ferrante, Laboratory of Socioecology and Social Evolution, Zoological Institute, Leuven, Belgium & Middlesex University, Dubai
  • Antonio Franchi, Laboratoire d’Analyse et d’Architecture des Systèmes, Toulouse, France
  • Evert Haasdijk, Vrije Universiteit, Faculteit der Exacte Wetenschappen, Amsterdam, Netherlands
  • Stéphane Puechmorel, Ecole Nationale de l’Aviation Civile, Toulouse, France
  • Nicolas Schneider, Airbus, Campus Engineering D42, Toulouse, France
  • Melanie Schranz, Lakeside Labs GmbH, Klagenfurt, Autriche
  • Clément Sire, Laboratoire de Physique Théorique, CNRS & Université de Toulouse (UPS), Toulouse, France
  • Gábor Vásárhelyi, ELTE Department of Biological Physics, Budapest, Hungary
  • Stéphane Viollet, Institut des Sciences du Mouvement Etienne-Jules Marey, CNRS & Aix-Marseille Université, Faculté des Sciences du Sport, Marseille, France


Click on the title of the presentations to read the abstract.

Monday, November 13th
Session 1: Decentralized control algorithms in swarms

Welcome Address – Guy Theraulaz & Bertrand Jouve


Bio-inspired collective behaviour of autonomous outdoor drone swarms – Gábor Vásárhelyi

Gábor Vásárhelyi
Department of Biological Physics
Eötvös University in Budapest, Hungary

Title: Bio-inspired collective behaviour of autonomous outdoor drone swarms

Abstract : While individual drones have gone through a tremendous development in the last years towards autonomy and intelligent behaviour, functional drone swarms are still very limited in number, due to the new levels of complexity a multi-agent system brings into the picture. The straightforward approach to multi-drone systems builds on the mindset of high individual intelligence and tries to increase the number of drones gradually. Contrarily, at our Department of Biological Physics at Eötvös University, we investigate large natural multi-agent systems and create statistical physical models optimized for the minimally required intelligence to perform a given task collectively. With this methodology, the typical bottlenecks of multi-agent systems are in the focus and thus there are less unexpected dynamical issues emerging when system size is increased. With our self-organized, distributed approach, we were able to recreate most of the basic building blocks of swarming behaviour both in simulation and on actual outdoor drone swarms of dozens of agents: synchronized collective motion, self-organized formation flights, coordinated autonomous drone traffic or collective object avoidance. In this talk I will give an introduction to our modelling concept and show our various results performed by our autonomous drone fleet.


Social interactions and collective states in fish schools – Clément Sire

Clément Sire
Laboratoire de Physique Théorique
CNRS, Université Paul Sabatier, Toulouse, France

Title: Social interactions and collective states in fish schools

Abstract : The flexible coordination of individuals’ movements ensures rapid and coherent changes in direction of travel of fish schools for instance as a reaction to a predator detected in the neighbourhood. However the ‘microscopic level’ interaction rules involved in the coordination of fish movements and the adapted collective response of a school still remain to a large extent unknown. Knowing such interaction rules could offer new sources of inspiration to design distributed control algorithms for swarms of drones. Here we present a systematic methodology to measure and analyze social interactions controlling the collective motion of animal groups. Contrary to classical forces between physical objects, social interactions between individuals explicitly depend on their relative headings and are affected by their anisotropic and asymmetric perception of their environment. Hence they strongly break the Newtonian’s law of action-reaction. When applied to fish groups, this approach leads to the quantitative measurement of the spontaneous behaviour of a fish, of its avoidance interaction with the tank walls, and of its attraction and alignment interaction with another fish. We use the results of this analysis to build an explicit and faithful model that convincingly reproduces quantitative and qualitative features of the actual fish dynamics. We also show that the type of models derived from such analysis reproduces the main collective states observed in actual fish schools, when one varies the intensity of the alignment and attraction interactions between fish.


Minimal models of collective motion from the engineering, statistical physics, and biological perspective – Eliseo Ferrante

Eliseo Ferrante
Middlesex University
Laboratory of Socioecology and Social Evolution
KU Leuven, Leuven, Belgium

Title: Minimal models of collective motion from the engineering, statistical physics, and biological perspective

Abstract: Swarm robotics studies the design of collective behaviours for swarms of robots by the development of controllers that only use local sensing and communication. One of the fundamental questions we ask in swarm robotics is how can we reduce the individual robot capabilities as much as possible and still obtain the desired collective behaviour. This is very important if we want to use robots with limited sensing capabilities, such as drones with payload restrictions. In this talk, I will describe a minimal swarm robotics model we developed to reproduce coordinated motion behaviour as seen in birds and fish.

This model required robots equipped with sensors able only to detect the relative position and angle of neighbours, and did not require the relative orientation of neighbours as in the classical Reynolds model. We showed that it achieved collective motion even in cases when no robot had a preferred direction of motion. Subsequently, we converted this mechanism into a novel statistical physics model which we called AES (Active Elastic Sheet). AES is based only on attraction-repulsion interactions as opposed to alignment-only interactions that characterize the standard statistical physics model of collective motion (the Vicsek model). In contrast with the Vicsek model, in a follow up work we also showed that AES is able to reproduce the same type of scale-free correlations as observed in natural starling flocks.

I will conclude by my talk by presenting some more recent developments and ideas, hoping to encourage joint projects with the participants to the workshop.


From fly to robots and vice versa – Stéphane Viollet

Stéphane Viollet
Institut des Sciences du Mouvement Etienne-Jules Marey
CNRS & Aix-Marseille Université, Marseille, France

Title: From fly to robots and vice versa

Abstract: The Biorobotic approach is a meeting point where robotics and neuroscience are used to try to explain the behaviour of animals, especially winged insects (fly, bee, wasp ..) and to model the processing of the sensory modalities used by these outstanding animals. The neurophysiology is also used to better understand the sensorimotor reflexes at work in the animals. For example, recent studies carried out at our laboratory focused on the hyperacidity in fly, i.e., the ability to locate a feature with a much greater accuracy than the one imposed by the optics.  Could future robotic applications take a great benefit of this bio-inspired approach? Several fly-inspired visuals sensors as well as autonomous bio-inspired robots will be presented in this talk.


Self-structuring of UAV traffic based on flocking behaviour – Daniel Delahaye & Stéphane Puechmorel

Daniel Delahaye & Stéphane Puechmorel
Dynamical systems Optimization team
Ecole Nationale de l’Aviation Civile, Toulouse, France

Title: Self-structuring of UAV traffic based on flocking behaviour

Abstract: UAV traffic management (UTM) is a very active area of research in which the major stakeholders of the air traffic system are involved. While air flow management is well established for civil aviation, it has yet to be defined for UAVs. In the Metropolis project, funded by the European commission and dedicated to autonomous vehicles, some possible structures for the flows in the airspace were investigated are compared from a capacity point of view. Based on its conclusions, it is appealing to use the same kind of organization within the frame of UTM. Due to the lack of a centralized control system, a fully distributed traffic management system is mandatory. In this presentation we will show how local rules of flocking may yield global structures similar to the ones introduced in Metropolis, namely streams organized in tubes, sheets or cells. Cohesion is ensured by enforcing a locally harmonic or biharmonic velocity field, in the spirit of usual flocking behaviour. However, to enforce the desired structure, the Laplacian used will be associated to an adhoc riemanian metric. Rules for aggregation and splitting are based on distance to central line of the target stream and distance to arrival. An example of such a self-organization will be presented on a simple swarm of quadrirotors.


Round Table 1

Tuesday, November 14th
Session 2: Mastering uncertainty and multi-level robustness in swarms

Collective Choice in Robot Swarms: The Best-of-n Problem – Marco Dorigo


Graph-Theoretical and Passivity-based Distributed Algorithms for Multi-robot Motion Control – Antonio Franchi

Antonio Franchi
Laboratoire d’Analyse et d’Architecture des Systèmes
CNRS, Toulouse, France

Title: Graph-Theoretical and Passivity-based Distributed Algorithms for Multi-robot Motion Control

Abstract: In this talk I will present a set of recent algorithms based on algebraic graph theory and passive system theory that allow group of robots to control their overall motion in a distributed way. Algebraic graph theory is used in order to model the robot interactions and communication, as well as to model fundamental properties such as graph connectivity and graph rigidity. Passivity theory is used in order to guarantee the stability of the overall system despite switching interconnections and estimation errors. The interconnection between graph theory and passivity in multi-robot system will be made clear by using the port-Hamiltonian formalism. I will conclude the talk with some real examples and experiments such as distributed target visiting with continual connectivity maintenance exploration with group of aerial robots.


Combining Environment-Driven Adaptation & Task-Driven Optimisation in Embodied Evolution – Evert Haasdijk

Evert Haasdijk
Computational Intelligence Group
Vrije Universiteit, Amsterdam, Netherlands

Title: Combining Environment-Driven Adaptation and Task-Driven Optimisation in Embodied Evolution

Abstract: Ficici et al. (1999) coined the phrase « embodied evolution » for evolutionary processes that are distributed over the robots in the population to allow them to adapt autonomously and continuously. Embodied evolution offers a unique opportunity for autonomous on-line adaptivity in robot collectives. The vision behind embodied evolution is one of collectives of truly autonomous robots that can adapt their behaviour to suit varying tasks and circumstances. Autonomy occurs at two levels: not only do the robots perform their tasks without external control, they also assess and adapt –through evolution– their behaviour without referral to external oversight and so learn autonomously.

Through embodied evolution robots adapt in their task environment; this implies that robots must adapt to their environment as well as learn to perform their user-defined task(s). I will discuss recent advances made researching the interaction of environmental and task-defined selection pressures in embodied evolution with the MONEE system (Haasdijk et al., 2014).


A Scalable Control and Estimation Framework for Robotic Swarms in Uncertain Environments – Spring Berman

Spring Berman
Autonomous Collective Systems Laboratory
Arizona State University, Tempe, USA

Title: A Scalable Control and Estimation Framework for Robotic Swarms in Uncertain Environments

Abstract : Robotic swarms are currently being developed to perform tasks over large spatial and temporal scales.  We are addressing the problem of reliably controlling swarms in scenarios where the robots lack global localization, prior data about the environment, and reliable inter-robot communication. As in natural swarms, the highly resource-constrained robots would be restricted to information obtained through local sensing and signaling.  We are developing a rigorous control and estimation framework for swarms that are subject to these constraints. This framework will enable swarms to operate largely autonomously, with user input consisting only of high-level directives that map to a small set of robot parameters.  We use stochastic and deterministic models from chemical kinetics and fluid dynamics to describe the robots’ roles, task transitions, spatiotemporal distributions, and manipulation dynamics at both the microscopic (individual) and macroscopic (population) levels.  In this talk, I will describe our work on various aspects of the framework, including stochastic strategies for coverage, feature mapping, and scalar field estimation, as well as decentralized, ant-inspired approaches to cooperative manipulation.  We are validating these techniques on small customizable robots, called “Pheeno,” that we have designed to be low-cost, versatile platforms for multi-robot research and education.


Bottom-up approaches and swarm behaviors to build robust and adaptive systems – Nicolas Schneider

Nicolas Schneider
Airbus, Toulouse, France

Title: Bottom-up approaches and swarm behaviours to build robust and adaptive systems

Abstract: As Airbus systems becomes more and more complex and in the mealtimes should evolves in more uncertain and unknown environments: it is obvious that next generation of flying machines should have the capabilities to behave and take decision in unexpected situations. Some research activities in Airbus are focused on the emergence of robust and adaptable behaviours for a large range of situations rather than designing very specifics systems that excel in very few operational contexts. Using bottom-up approaches and the emergence of behaviours from collaborative systems, we have developed several prototypes, using flying drones and rolling robots to address this problem. During the presentation we will go through some examples and shared expectations / limitations of such approaches from an industrial point of view.


CPSwarm: A Horizon 2020 Project on System Integration and Tools to Support Engineering of CPS Swarms – Melanie Schranz

Melanie Schranz
Lakeside Labs GmbH
Klagenfurt, Austria

Title: CPSwarm: A Horizon 2020 Project on System Integration and Tools to Support Engineering of CPS Swarms

Abstract: Cyber-Physical Systems (CPS) find applications in a number of large-scale, safety-critical domains e.g. transportation, smart cities, etc. While the increased CPS adoption has resulted in the maturation of solutions for CPS development, a single consistent science of system integration for CPS has not yet been consolidated. Therefore CPS development remains a complex and error-prone task, often requiring a collection of separate tools. Moreover, interactions amongst CPS might lead to new behaviors and emerging properties, often with unpredictable results. Rather than being an unwanted byproduct, these interactions can become an advantage if explicitly managed at early design stages.
CPSwarm tackles this challenge by proposing a new science of system integration and tools to support engineering of CPS swarms. CPSwarm tools will ease development and integration of complex herds of heterogeneous CPS that collaborate based on local policies and that exhibit a collective behavior capable of solving complex, industrial-driven, real-world problems.
The project defines a complete toolchain that enables the designer to:

  • set-up collaborative autonomous CPSs;
  • test the swarm performance with respect to the design goal; and
  • massively deploy solutions towards “reconfigurable” CPS devices.

For more information visit our website: http://www.cpswarm.eu


Round Table 2

Follow-up action plan :

Following the workshop, the considered action plan should include :

–       A call for proposals to a special edition of Swarm Intelligence on distributed control of swarms of drones (4-5 articles)

–       A exploratory project financing masters internships on 1) Control of swarms: adaptation, decentralized decision and local implementation; and / or 2) Mastering uncertainty in swarms: local/global perception, pattern formation and evolution
–       A interdisciplinary project in a coming call for proposals

If you are interested to join, please kindly notify us by sending a message to Guy Theraulaz (guy.theraulaz @ univ-tlse3.fr) and Aurélie Cretté (crette.aurelie @ gmail.com) and we will share further steps as to how we can collaborate in future.

Pictures :

20171113 152558
Eliseo Ferrante
Gabor Vasarhelyi
Clément Sire
Stéphane Puechmorel
Marco Dorrigo
Spring Berman
Evert Haasdjik
Melanie Schranz
DOhjl1IXkAIex K

Contacts :

  • Guy Theraulaz : guy.theraulaz@univ-tlse3.fr
  • Aurélie Cretté : crette.aurelie@gmail.com

The SYSCOB Project


Funded by :

Supported by :