Visit Your Local PBS Station PBS Home PBS Home Programs A-Z TV Schedules Watch Video Donate Shop PBS Search PBS
SAF Archives  search ask the scientists in the classroom cool science
scientists from previous shows
cool careers in science
ask the scientists
SCIENTISTS FROM PREVIOUS SHOWS

Photo of Thiemo Krink Thiemo Krink as seen on Spiders!: Spin, Spin, Spin...

Click on Thiemo's photo to read a brief bio.







q I am wondering if you have your virtual spiders program available for the public? (asked by many viewers)

A Currently, we have a version for the Macintosh. However, we haven't decided if this version should be public yet. Please let me know if you would like to have a copy.



q I listened to your description of your digital spider and I understood that you are using a Genetic Algorithm. I'm interested in reproducing such a digital spider program. The only thing I'm not sure of is what genes are simulated in your program. You say that only two parameters govern the behavior of the digital spiders (angle and spacing of the web strands, I think...) but the webs displayed seems very intricate for only two parameters. Can you describe in more detail how the simulated genes are used to create the digital spider web? Pierre

A There are in fact more rules and parameters involved, though the system is still very simple. However, the main factors which distinguish the shape of a capture spiral made by a garden cross spider from other orb weaving spiders can be simulated by two simple rules. In the show, two different projects on spider modeling have been introduced: "Theseus" is a simulation of a virtual spider robot, which completes the half-done web of a real garden cross spider. In "NetSpinner," virtual spiders have a set of parameter controlled rules to build entire virtual webs. These parameters are encoded as artificial genes, which are recombined when spiders mate and transferred to future generations. These genes control parameters such as the number of start radii for a web, the average spacing between successive turns of the capture spiral, and a number of other variables. Here, the objective was to investigate the effect of different environmental scenarios on the evolution of spiders' webs, such as varying sizes of prey items, escape strategies of captured prey, spatial constraints of the web-building site.

Please find more detailed information in:

Papers:

Krink, T. & Vollrath, F. 1997. Analysing spider web-building behaviour with rule-based simulations and genetic algorithms. In: Journal of Theoretical Biology, 185, pp. 321-331.

Krink, T. & Vollrath, F. 1998. Emergent properties in the behaviour of a virtual spider robot. In: Proceedings of the Royal Society, 265, pp. 2051-2055.

Krink, T. & Vollrath, F. 1999. Using a virtual robot to model the employment of regenerated legs in an orb spider. In: Animal Behaviour, 57, pp. 223-241.

Books:
Climbing Mount Improbable, Dawkins, Richard, chapter 2, pp. 51 - 60, 1996, Penguin Books, London

Scientific Magazines:
The Spider's Stratagem, New Scientist, cover story, No. 2042, pp. 24-28, 1996

Websites:
My web page: www.brics.dk/~krink/





q How does the information that a virtual parent spider learns from trial and error get "coded" on its DNA to be passed down to the next generation? Patrick

A This process is based on a simplified model of natural evolution. At simulation start-up, we create a population of virtual spiders with random genes. The spiders which build the "best" webs, i.e. webs with a good prey catching success and minimal need for silk, are selected to become the parents of the next generation of virtual spiders. In this process, there artificial genes are recombined (by a process called crossing over) and mutated. Conclusively, "bad" ideas will get lost when individuals die without reproducing and good ideas will spread when virtual spiders have offspring.



q Do you think the web-building skills of your virtual spiders will ever surpass those of real spiders? Joshua

A Real spiders have to deal with far more problems than our virtual spiders. A real spider also needs to manage different life stages, search mates, fight with other predators, deal with parasites, etc. In addition, there are a lot of environmental factors which we have to simplify in the computer model. Further, the real world is a dynamic environment with a lot of fast and unpredictable changes, which makes life difficult for a real spider. Therefore, the web is a compromise between different needs. Thus, the virtual spiders might be able to build webs which are better to catch artificial prey than the real spiders, but there task is still quite different.



q question here

A What sort of web building rules do you give the Cyber Spiders? Steph

The rules and behaviour patterns that controlled "Theseus" actions (which is the virtual spider that completes the half-done digitized web of a real spider) were based on extracted hypotheses from (video-taped and analyzed) observations of web-building behaviour. Rules and behaviour patterns for capture spiral construction were organized in three groups: (1) basic orientation and movement, mainly during (i) searching for radials or peripheral frame and spiral threads, (ii) orientation along the auxiliary spiral and (iii) u-turn behaviour to maintain capture spiral spacing; (2) web manipulations like fixing new sticky threads and removing auxiliary threads when they are no longer needed for support or orientation (the exact location of a joint is determined by a rule compromising between (i) keeping a constant distance to the currently tracked auxiliary spiral thread and (ii) keeping a constant distance between successive turns of the capture spiral); finally, (iii) gravity as a secondary spatial information cue for fine adjustments according to the spider's slower prey attack speed when running upwards.

For instance, one rule of category (1) (i) tells the spider to search for an intersection between an auxiliary spiral thread and a radius with the foreleg that it currently uses for orientation guidance.

The complete list of rules for "Theseus" is described in:


Krink, T. & Vollrath, F. 1999. Using a virtual robot to model the employment of regenerated legs in an orb spider. In: Animal Behaviour, 57, pp. 223-241.

The list of rules for "NetSpinner" (the evolution of spider webs) is described in:
Krink, T. & Vollrath, F. 1997. Analysing spider web-building behaviour with rule-based simulations and genetic algorithms. In: Journal of theoretical Biology, 185, pp. 321-331.




 

Scientific American Frontiers
Fall 1990 to Spring 2000
Sponsored by GTE Corporation,
now a part of Verizon Communications Inc.