Description of LEURRE

Lures in LEURRE and Artificial Life control

The use of lures is very old in the history of interactions between humans and animals. These lures are often the result of a tradition, obtained by “trials and errors”. Most of the time, previous studies of the behaviors of the animal (isolated or in collectivities) are absent. Moreover, these lures do not interact together and do not have any adaptive capability.
At the individual level, behavioral sciences have shown that animals’ interactions could be rather simple signals and that it is possible to interact with animals not only by mimicking their whole behaviors but also by making specifically designed artifacts. At the social level, the theory of self-organisation applied to animal societies shows that very simple, but numerous, interactions taking place between individuals may ensure complex performances at the level of the group and produce collective capabilities. In this context, LEURRE will develop mixed-societies composed of animals and robots that interact and communicate.
One of the main fundamental questions in social biology is to understand the link between individual and collective behavior. An elegant way to identify individual behavioral algorithms consist in replacing some animals within a group by robots and comparing collective responses in “mixed" and “natural" groups.
The main part of LEURRE will deal with experiments blending together cockroaches and small insect-like robots, the LEURRE insbots. The main goal is to demonstrate that it is possible to mix insects and specifically designed robots that interact and communicate. We will then show that the lure robots allow the control of the global behavior of a mixed-society.
Another important part of the project is the experimental study of individual and collective behavior of sheep. Then, the project will use the gained methodology developed in the first part of the project to evaluate its application in agriculture, especially in sheep breeding where lures could be put at work to induce desired collective behavior at the level of the herd.

LEURRE project and complexity science

Societies (as groups of animals, of robots, or mixed-societies of robots and animals) are complex systems with non-linear dynamics. The project will be focused on classical questions in complex systems: studies of pattern formation i. e. their functionality and the mechanisms needed to create these patterns. Different disciplines interested in Complex Systems are present in the project: Physics with concepts like multi-stationnarity, spontaneous pattern formation and self-organisation; Ethology with the study of animal societies; Artificial Intelligence and Information Theory with distributed computing and distributed information and their consequences on pattern formation.
In societies (including mixed-societies), the information on the environment is changing and incomplete. In this context, behaviors and communication have distinctly random elements moreover signals can be blocked or can interfere with each other. Far from being undesirable, individual randomness can help the team to reach collective solutions that would otherwise be overlooked. Amplifying communication is a characteristic of group-living animals (e. g. recruitment, social imitation,... Most self-organized decisions and patterns arise as a result of a competition between different sources of information that are then amplified through different forms of positive feedback. In contrast, negative feedback often arises automatically as a result of the systems constraints (e. g. limits on the supply of food, the number of available individuals,...).

LEURRE and the Quality of Life

EU policy encourages research that will strengthen the technical and economic competitiveness of bio-based industries through integration of various scientific disciplines. This means the development of new and sustainable systems of production, including breeding methods and exploitation in agriculture, fisheries and aquaculture, taking into account profitability, the sustainable management of resources, product quality and employment as well as animal health and welfare.
Modern agriculture is looking for new methods, increasing the animal welfare and less disturbing for the animals and the environment. In the management of animal populations, it means to take account of the basic behavior of the animals. This multidisciplinary project fits with these aims and will provide the scientific and technological basis for the development of new innovative breeding and management methods which integrate basic and social behavior of the animals. Indeed, most of the species that we are breeding or fishing are social, living in large groups. It will contribute to the improvement of health and animal welfare and a better environmental management. LEURRE has also potential applications in natural systems like the management of the pest and invasive species which are often highly social.
Environment protection, animal welfare or animal feed safety and economic growth need to be reconciled. This reconciliation requires greater understanding of the environment and the living organisms and development of information sciences. Improving the way natural resources are managed is an important investment in Europe’s future. We need to understand biological systems better.
In the context of the control and the management of animal natural resources, the importance of the behavioral and social levels are often underestimate. LEURRE will show how these knowledge can be implemented and what are the methodology, the tools and technologies we need. Together with the conceptual knowledge developed in the project, social environmental organisers could use behavioral control to direct desired collective response.

LEURRE and its importance for Europe

The LEURRE project contributes to the Community’s research policy: LEURRE consortium brings together the different knowledge required to successfully meet the project objectives and establishing communication links for other interdisciplinary research in artificial systems and the life sciences.
The control of interactions between artificial systems and living organisms is thus a key aspect in the design of artificial systems, as well as in many agricultural, medical, scientific and technical fields. Moreover, different EU programs, highlight the growing importance of this synergy between the artificial and natural systems. The fields of collective intelligence and micro-robotics are today critical and both research and industry are urged to find tools en methods for understanding, and controlling, fast growing fields of the technology such as collective robotics. The collective intelligence (or self-organisation) in natural and artificial systems, bio-inspired robotics and micro-robotics and modelling of multi-agents systems are the subject of active research and Europe is already a leader in these fields.
LEURRE has the potential to put Europe in a leading role in these new fields which are the mixed artificial - natural systems and the collective (bio)robotics. The project is expected to bring advances in these fields and to give Europe a leading position in this new field of mixed societies and control of natural populations with autonomous and/or bio-mimetic agents.
Standardisation Though this is a long-term goal the project could contribute to standardisation. Standards are an important thing in the industry and science. They create markets that can either be closed (proprietary standards) or open (open standards). Open standards have shown to be able to create markets in which a wide range of diverse players can compete or find niches, generally to the benefit of the customer. In order to prepare for such an open market for the technologies of emergence control the consortium judged it to be important to set standards into place as quickly as possible. All the tasks related to the modelling and development of the behavioral algorithms are the roots of the standardisation.
This project prepares the achievement of the anticipated industrial and economic results by improving the basic approach to monitor and control blend societies where humans, animals and artefacts interact.
It is therefore of outmost important that the results of the project are disseminated to the scientific and industrial world to stimulate and catalyse the development of laboratory prototypes, industrial prototypes and spin offs. This will be realised by scientific articles by each of the partners in peer reviewed journals, but also by an easily available website and by workshops targeted to potential users of the results.

Objectives of LEURRE

The behavioral model

LEURRE will propose a modeling protocol to be applied to the calculus of emergence in mixed-societies and study its properties. Its goal is similar to research in collective intelligence or self-organization, namely to understand how phenomena at one level of biological organization (here social) emerge from the properties of lower-level units (individual behavior). Having determined the pathways of information flow among the units and their behavioral rules, analytical and numerical models enable to predict the properties of the mixed society.
LEURRE is mainly concerned with the dynamics and the emergence of collective patterns. We consider three types of entities: the animals, the robots and the chemical/physical variables (e. g. pheromone, sound, building material, resource, shelter,. . . ).
The robots and animals (the autonomous agents) are characterized by their position, orientation, activity and internal state (e.g. a physiological variable, memory,...). Each unit obeys rules that determine its reactions as a function of (i) the signals it receives from the environment and others individuals, (ii) their internal state, (iii) their present activity. The responses are probabilistic and lead to a change of their activity, internal state, position and movement. In turn, these changes affect the environment and the other natural or artificial entities of the group.
The chemical or physical variables are governed by rules describing their dynamics in space and time. For instance, in the case of pheromones, differential equations describing the time evolution of their concentrations are sufficient. These equations comprise three terms: (1) the "birth" (production); (2) the "death" (destruction, disappearance e.g through evaporation); (3) the propagation (diffusion, convection,...). Because we work with behavior and communications based on the intensity, gradient,... and with mixed societies, the robots and animals are governed by similar behavioral rules and communications.
It implies that the mathematical tools used to describe the behaviors of the two categories of autonomous agents are the same. These behavioral models are then formalized into programming languages to use them in computer simulations and to store them in a behavior database together with a library of tools (algorithms, constraints,...). A user-friendly interface will be provided for using this database and make computer simulations.

Experimental and computational realization of mixed-societies

The main goal of LEURRE is a real-life realization. The project aims to build mixed societies of living insects with specifically designed robots or insbots (insect-like robots).
The behavioral model is one of the tools to manage this realization. Firstly, we will develop models of several behaviors of insects (e. g. random-walk, probability to stop, probability to stay in a cluster,. . . ). Secondly, this model will be mapped into the insbots that will "inherit" the defined behaviors and protocols. The engineering of interactions between artifacts colonies and animal colonies (living organisms) is a key aspect of the design of artificial systems. Complex performances do not always require complex individual behavior. It is our hypothesis that interesting developments also pertain to simpler interactions.
The main problem in the LEURRE project is communication and interaction between animals and robots. The signal emitted by the animals has to be detected by the robot and, vice versa, the robot has to be able to emit signals detected by the animals. Hence, it is important to choose situations in which communication is not only simple, but also that the signals used by the animals to communicate are easily detected by the sensors of the robots. This is one of important limiting steps of the project. However, in many animal groups that display collective intelligence, the signal remains relatively simple to detect, being for instance, movement, position of the body, or simple chemical signals. This is why the gregarious cockroaches are our experimental model.
Controlling the global behavior of a mixed-society is feasible. Here, control means directing the emergence of some global pattern. We will show that it is possible to trigger such an emergence by temporarily or permanently adding to the system entities with specific behaviors. In the LEURRE project, it will be done by introducing insbots, as lures, in a group of insects. These insbots will have to be distributed (by the nature of the system), adaptive (because of the interaction with the animal colony) and robust (to assure control). To highlight the effective communication, we will show that a collective behavior, such as aggregation, can be controlled by the use of lures. The insbots will be able to push the group towards a totally new collective behavior such as an unnatural pattern of aggregation. Moreover, we will show (for some situations) that after reaching the new pattern, the lures can be removed without changing the collective behavior that remains stable.

Towards a modeling and experimental methodology to control mixed-societies

Guidelines towards a methodology for the study and the control of mixed-societies will be provided. One fundamental goal here is to understand how the superstructures arise from the actions and interactions of the members of the group: “What are the typically configurations that support a priori emerging organization of mixedsocieties”.
Hence the aim here is identical to that of much research in CI (Collective Intelligence), namely to understand how phenomena at one level of biological organization emerge from the properties of lower-level units. We need to determine the pathways of information flow among the units and their behavioral rules of thumb. Unfortunately, it is generally extremely difficult to predict, whithout a formal modeling, the properties of complex systems, such as animal groups, which are comprised of multiple components with complex, dynamic interactions.
However, mathematical equations and computers enable one to predict the properties of complex systems, and so they provide a means of evaluating one’s model of a group’s internal machinery. The approaches just described are likely to yield strong suggestions about how a group works, but testing the accuracy and completeness of one’s understanding requires a further stage in the analysis: formulating a rigorous model which embodies one’s current understanding of how the group works. Here one takes a bottom-up approach to model building, to give shape to the model.

Relevance of our results to quality of life and management of living resources

The evidence of our results to the control of emerging configurations in agriculture and environment will be provided. There are needs of new and sustainable systems of breeding methods and exploitation in agriculture and fisheries, taking into account profitability, as well as animal health and welfare. We will show how our methodology (quantification, modeling and use of intelligent agents to control group) and our results obtained with the model cockroach-insbot may be extended to these questions.
Most of the species that we are breeding of fishing show highly collective behavior. Despite, the differences between insects and mammals societies, similar basic rules govern the organization of these species. Many behavorial issues appearing in animal societies have a strong self-organised component: synchronisation of activities, aggregation, sorting,...
As most self-organised systems are very sensitive to small quantitative changes at the individual level, it is possible that a limited number of robots interacting within the group might give the opportunity to the animals to escape from some sub-optimal solution by introducing new collective behaviours. Breeding domestic animals, taking into account their welfare is concerned with such mixed groups of robots and animals. For instance, the control of the behaviour of domestic fowl and other birds of poultry farming gathered in very large number is one example. Social imitation plays a key role in these species and most of their collective behaviour results from positive feedback. An example of application could be the control of collective panic movements, which may result from such positive feedback and often induce high mortality in some species. A robot (or a group of robots) could be able to induce a new wanted behaviour or modify the organisation of the group in order to damp the snowball effect of collective panic.
In the LEURRE project, special attention will be paid on breeding and life management of sheep. A concrete problem we will tackle using our methodology is to model the collective behavior of sheep and their control with an artificial agent.
Another example concerns pests that are often highly social species. In general, methods of management acting at the level of the social behaviour should be much more efficient than the usual destructive methods. One example of pest management refers to birds roosting by thousands in a given place that might be the source of various environmental problems especially for local inhabitants. In such context, a control of the roosting behaviour by using relatively simple robots would be able to affect the spatial distribution of these wild animals.
Robotics has much to learn from ethology while robotics may surely help ethology to explore animal behaviour. The main goals of this project is to show that complex collective responses may emerge from individual simplicity and simple signals and that experiments using lures mimicking these simple signals are able to modify the group behaviour. The different problems to solve and the potential benefits that can be gained by mixed societies are sufficiently important to initiate and develop research in this direction.

Realization of LEURRE

The consortium

The LEURRE project involves the combined effort of the following partners: the Université Libre de Bruxelles (ULB, Brussels), co-ordinator of LEURRE, the International Solvay Institutes for Physics and Chemistry (ISI, Brussels), technical leader of LEURRE, the Université de Rennes I with the French CNRS (EVE, Rennes), the Université Paul Sabatier with the French CNRS (UPS, Toulouse), the Ecole Polytechnique Fédérale de Lausanne (EPFL, Lausanne) and one subcontractor A. Martinoli (Lausanne). The teams of the consortium are familiar with social behavior, collective intelligence, complex and artificial systems. They are also highly complementarily, with people who share world-leading competence into the domains of investigation concerned with the project. The teams are specialists of ethology and animal societies (ULB, UPS, EVE), models in the fields concerned with the project (ISI, EPFL, ULB) analysis of collective behavior (ULB, UPS), individual and collective robotics (EPFL). Broadly speaking, the teams are specialists of identification of the mechanisms and their adaptive value or they have competence and experience in building and managing artificial systems. Moreover — and it is important for this interdisciplinary project — these teams have common interest, complementary knowledge and some tradition of collaboration. All partners have previous experience with interdisciplinary work (e. g. experimental and theoretical works; biomimetics project) and most of the partners have already collaborated successfully before in a European context. The project can be described as a strong web of interactions between the teams. Each team is involved, at least, in two tasks. Moreover for each task, at least, two teams will co-operate. The project needs an European level and cannot be conducted at the national level for both reasons of complementary knowledge and critical mass.

• EPFL-ASL1: with its experience in building autonomous robots. This team is responsible for the design and the building of the insbots. Their first contribution will be in work-package 1. Their second contribution will be in workpackage 2. They will be also involved in work-package 3.
• EPFL-SWIS: with its experience in development of (automatic) design, modeling, control, and optimization methodologies for self-organized, collectively intelligent, distributed systems. Its contribution will be in work-package 1 & 2.
• EVE: a specialist of cockroach ethology, will be highgly involved in the design of the Insbots as well as the tests between insbots and insects. The first work-package for EVE will be work-package 1, but EVE will also contribute to work-package 2.
• ISI: the research activities of the Solvay Institutes involve both fundamental and applied problems in non-linear science. The ISI will contribute to apply innovative techniques to implement the behavioral algorithms and to design the behavior database (work-package 1). ISI and the ULB are responsible of the different theoretical tools (simulations) of work-package 2 and workpackage 3.
• ULB with its experience in collective behavior and ethology is responsible for the different collective experiments with the help of the UPS (for his experience of image analysis and data analysis applied to the collective behavior in work-package 2 and work-package 3) .
• UPS with its experience in mammals behavior is responsible of the task Agriculture (work-package 3). UPS will also contribute to the behavioral modeling (work-package 1) and to the image and data analysis package.

All partners will contribute to the project initialisation, the evaluation and the tests of results and to the dissemination of outcome, each from their own point of view. However, ISI will undertake major effort in the dissemination of the outcome to stimulate new applications.
The roles of the partners complement each other; on the other hand, the consortium is complete with respect to the objectives sought. Moreover, with the planned work programme, close co-operation is ensured: the continuous interplay between tasks and exchanges of research outcomes and intermediate deliverables will be the key to the success of the project.

top

News