We wish to design a general methodology for designing mixed animal-robot societies, in which the robots can control the animal swarm/herd/flock in a desired away. If successful, we can take advantage of sensing, actuation or perception capabilities of animals that are hard or expensive to reproduce with engineering systems. For example, we could control sheep flocks to mow our lawns, bird flocks to gather environmental data or rats to search for explosives.

Collective decisions in Cockroach Swarms

A mixed natural-artificial society of cockroaches and robots

The project Leurre was sponsored by the Information Society Technologies program of the European Community and aimed at the development of a general methodology to control animal societies by introducing animal-like robots by experimenting with cockroaches and cockroach-like robots. The primary goal of the project was to show that introducing cockroach robots into a cockroach swarm can influence collective decisions made by the swarm.

For this, we presented a swarm of cockroaches with two shelters, a darker one and a brighter one under which the insects can aggregate. Naturally, the insects almost always preferred the darker over the brighter shelter. As they are gregarious animals, these decisions are made collectively and the insects prefer to stay together. We then programmed a swarm of miniature robot Insbots that were impregnated with cockroach  pheromone to aggregate under the brighter shelter. Interestingly, adding only a few robots that exhibit this non-natural behavior was sufficient to significantly alter the behavior of the natural swarm to join the insbots under the bright shelter.

While experiments involving mixed societies of american cockroaches and Insbots are conducted at ULB, we were concerned with developing models on various levels of abstractions (from realistic simulation to macroscopic difference equations) that help to understand the dynamics of aggregation and collective choices.

Cow Herding with Virtual Fences

Cow-herd equipped with GPS-enabled helmets

We are also interested in distributed controllers that can contain cattle within a moving virtual fence line. Researchers around Daniela Rus at MIT developed a device that can be mounted on a cow’s head and apply a stimulus to the cow as a function of its GPS location. These devices have also been used for identifying a behavioral model for the individual cows, which is the basis for our study.

We study the following question: assuming that inducing stress increases the gregarious behavior of the stressed animal, are undirected cues sufficient for containing the herd in a moving perimeter?

For investigating this problem, we add three parameters to the behavioral model: stress propagates to all cows within a certain radius R, stress decays after an average time T, and the gregarious behavior during stress is increased by a factor alpha.

Results show that if mutual attraction between animals remains the same during stress, undirected cues are useless. If we assume that mutual attraction increases during stress, however, the herd can be successfully controlled by controlling only a fraction of the animals.

Publications

F. Mondada, A. Martinoli, N. Correll, A. Gribovskiy, J. Halloy, R. Siegwart and J.-L. Deneubourg. A General Methodology for the Control of Mixed Natural-Artificial Societies. In Handbook of Collective Robotics, Serge Kerbach, editor, Pan Stanford Publishing. To appear.

N. Correll, M. Schwager and D. Rus. Social Control of Herd Animals by Integration of Artificially Controlled Congeners. In the 10th International Conference on Simulation of Adaptive Behavior, Osaka, Japan, 2008. Springer Lecture Notes in Artificial Intelligence (LNAI) 5040, pages 437-447. Best Paper Award.

J. Halloy, G. Sempo, G. Caprari, C. Rivault, M. Asadpour, F. Tache, I. Said, V. Durier, S. Canonge, J. Ame, C. Detrain, N. Correll, A. Martinoli, F. Mondada, R. Siegwart, and J.-L. Deneubourg. Social Integration of Robots into Groups of Cockroaches to Control Self-Organized Choices. Science, 318(5853):1155-1158, 2007. [preprint]

Links

Distributed Intelligent Systems and Algorithms Laboratory at EPFL

 

 
  • Pingback: Animal robotics | Greenovationhawaii

Set your Twitter account name in your settings to use the TwitterBar Section.