Evolving Gene Regulatory Networks for Real Time Control of Foraging Behaviours - supplementary materials

This page contains videos that supplement figures for the 12th International Conference on the Synthesis and Simulation of Living Systems (ALIFE XII) paper by Michal Joachimczak and Borys Wrobel.
We are in the Evolutionary Systems Lab, Institute of Oceanology, Polish Academy of Sciences.
If you have any questions or comments, don't hesitate to contact me: mjoach ( at ) gmail . com

Contents:

  1. Abstract
  2. Figure 6
  3. Figure 5
  4. Figure 12
  5. Figure 11
  6. Reference/BibTeX
  7. Previous work with the same system

1. Paper abstract

We use a genetic algorithm to obtain artificial gene regulatory networks (GRNs) controlling real time behaviour of artificial agents (animats) that gather food resources in a 2D environment. We build a system in which evolving GRNs are encoded in linear genomes. The encoding allows to determine which transcriptional factors (TFs) interact with which regulatory regions (promoters) to form a GRN. The sensory information is provided to an animat as externally driven concentration of selected TFs. Concentration of selected internally produced TFs is interpreted as signals for actuators. We first consider foraging for one food source and then scale the problem up to obtain animats that are able to switch between two types of food sources and avoid the poisonous one. We show that our system is highly evolvable, even though the genome encoding is very flexible (which results in a large search space) and though continuous product accumulation and degradation causes latencies in signal processing by the networks. We then discuss the topological properties of evolved networks and their evolutionary trajectories. Our results provide a first step toward a more ambitious goal of developing an artificial ecosystem in which multiple individuals will compete for food and mates.



2. Figure 6 - best individual for single type of food source

This figure demonstrates behaviour of the best individual evolved to seek single type of food particles. Right panel shows the scent intensity that is locally perceived by the two of animat sensors (drawn as small circles on the front of the animat). The force applied by the actuators is drawn on the sides of the animat as a line of proportional length.
Behaviour of the same individual on two different maps is shown.



Map #1 (same as in the paper)
Map #1 (with scent intensity, 0-blue, max-red)



Map #2
Map #2 (with scent intensity, 0-blue, max-red)



3. Figure 5 - common local minima in the search space, suboptimal, circular behaviour

This behaviour relies on consuming the food particles by performing circular motion. Despite low average speed, the food is being actively searched for.


Map #1 (same as in the paper, rotated)
Map #2 (same individual, different map)



4. Figure 12 - best individual for the problem with two food sources, generation 5000

Initially, the red particles are poisonous. After consuming 5 blue particles, the switch occurs, and the blue ones become poisonous. This setting thus requires the GRN to evolve separate control modules and apply seeking/avoidance adequately.

Map #1 (same as in the paper)
Map #2 (same individual, different map)



5. Figure 11 - best individual from the same run as in Figure 12, generation 2600

Again, the evolution goes through the stage of circular behaviour, similar to that seen on Figure 5 in the paper. Also, the speed is greatly improved over next 2400 generations.

Map #1 (same as in the paper)
Map #2 (same individual, different map)



6. Reference/Bibtex

Joachimczak M., Wróbel B. (2010) Evolving gene regulatory networks for real time control of foraging behaviours In: Artificial Life XII: Proceedings of the Twelfth International Conference on the Synthesis and Simulation of Living Systems (Edited by: H. Fellermann, M. Dörr, M.M. Hanczyc, L.L. Laursen, S. Maurer, D. Merkle, P.-A. Monnard, K. Stoy, S. Rasmussen), pages 348-355 MIT Press, Boston, MA.
BibTeX / RIS /  CiteULike 


7. Previous work on the same system

Have a look at our earlier work on the same model of gene regulatory network, applied to 3D development and patterning. If you can't access the pdf or it's not yet available, just e-mail me (see top of the page).

Evolution of the morphology and patterning of artificial embryos: scaling the tricolour problem to the third dimension. In: Proceedings of 10th European Conference on Artificial Life (ECAL 2009), volume 5777 of LNCS, pages 33-41, Springer.
BibTeX / RIS /  CiteULike 
Videos from the paper are available.


M. Joachimczak and B. Wrobel. Evo-devo in silico: a model of a gene network regulating multicellular development in 3D space with artificial physics. In S. Bullock, J. Noble, R. Watson, and M. A. Bedau, editors, Artificial Life XI: Proceedings of the Eleventh International Conference on the Simulation and Synthesis of Living Systems, pages 297-304. MIT Press, Cambridge, MA, 2008. BibTeX / RIS /  CiteULike
Videos from the paper are available.


We have also explored evolvability of this GRN model to the problem of processing changing external signals and simple computation tasks in a parallel paper:

M. Joachimczak and B. Wrobel. Processing signals with evolving artificial gene regulatory networks. In Artificial Life XII: Proceedings of the Twelfth International Conference on the Simulation and Synthesis of Living Systems. MIT Press, Cambridge, MA, 2010. (in press) BibTeX / RIS /  CiteULike



Last modified: July 2012