Formalisation of reactive multi-agents systems
Scientific problem
Our research aims at providing a new description of multi-agent systems that would separate, on the one hand, the description
of the agents, and, on the other hand, the simulation scheme.
By ``simulation scheme'', we mean all the operations that regard the
interactions between agents and between agents and the environment.
We consider such a separation as a means of answering two questions:
- How to describe a multi-agent system with a mathematical formalism
that removes ambiguities of formulation and that allows readers to
reproduce an experiment with precision?
- To which extent does the global behaviour of a multi-agent system depend on the laws governing the agents and to which extent does it depend on the
scheme used to run a simulation?
Turmites
Turmites, also called Langton's ants [Lan86],
are very simple agents that evolve on a grid.
The grid is constituted of cells, which can be in two states 0 (white) or 1 (blue).
The rules govering the agents are:
- If a Turmite is a 0-cell, it turns it to a 1 and turns left and moves forward
- If a Turmite is a 1-cell, it turns it to a 0 and turns right and moves forward
We take a mutli-Turmite model as an example to illustrate our method.
[Lan86] Christopher G. Langton, Studying Artificial Life with
cellular automata, Physica D, n° 22 (1986)
Article
Multi-agent Systems as Discrete Dynamical
Systems: Influences and Reactions as a Principle
Download preprint
Logiciel
Clik here
to download FiatLux and
test the multi-Turmite model under different simulating schemes.
Launch with a double-click. Press on tab "Multi-agents" and then press "Go" to test for different configurations. Please write to us if you need more detailed information.
Simulations
We provide to the readers of the article some animations that correspond to the different exepriments from the paper.
Click on the image to display the film in full screen.
Please press the "Reload" button of your browser to reinitialise the film diplay.
Figure number (in article) | initial configuration |
Comments | Film step |
1 (left) |
|
One single turmite |
100 |
1 (right) |
|
One single turmite with north cell flipped |
100 |
2 (left-middle-right) |
|
Four Turmites : short time evolution |
1 |
2 (left-middle-right) |
|
Four Turmites : long time evolution |
100 |
3 (left-middle-right) |
|
Two Turmites : short time evolution |
1 |
3 (left-middle-right) |
|
Two Turmites : long time evolution |
100 |
4 (top) |
|
Evolution of two Turmites |
1 |
4 (middle) |
|
Evolution of two Turmites |
20 |
4 (bottom) |
|
Evolution of two Turmites |
4 |
5 |
|
Deadlock with two Turmites |
1 |
8 |
|
Observation of two gliders |
5 |
For other figures, please use the FiatLux simulator.
Feel free to contact us if you have any questions.
|