Three dimentional spider webs research
Robotic fabrication of an architectural element proposal


This project is focusing on biomimmetic analysis of a three dimensional spider webs and possibilities for implementation of the research results into an architectural robotic fabrication process. We were mostly interested in investigating web construction process. How does the spider build such complex and stable structures? We analyzed its movement during the web building process using photogrammetric methods. We have discovered two different types of behavior and implemented it into a fabrication process that consists of two stages. (continue below images)

raw data.jpg
physical 2.jpg
Stuttgart University 2014
ITECH Masters program
Tutors :
Prof. Achim Menges, D. Correa, M. Dörstelmann, M. Prado
Prof. Jan Knippers, V. Koslowski, S. Parascho, G. Schieber
Prof. Dr. James Nebelsick, C. Allgaier
Maria Yablonina
Paul Pionet
Yiliya Baranovskaya
Yassmin Al-Khasawneh


What is a spider web?


A spider web (or cobweb) is a device created by a spider out of proteinaceous spider silk extruded from its spinnerets. 

Most spiders have three pairs of spinnerets, each having its own function – there are also spiders with just one pair and others with as many as four pairs. Webs allow a spider to catch prey without having to spend energy on running it down.

The tensile strength of spider silk is greater than the same weight of steel. Its microstructure is under investigation for potential applications in industry, including bullet-proof vests and artificial tendons. Researchers have used genetically modified mammals to produce the proteins needed to make this material.


Tegenaria Attica

Finding Diane


From the very beginning of the research on spider webs we agreed to not concentrate on the material of the web, but look into the construction process of a three dimensional web. As far as there is not much research done on that topic in biological world, we decided to try and analyse it ourselves.

First thing we would need for that would be a spider. We had to find a species that would match all our re­quirements. First of all, it has to be a spider that pro­duces a web with complex multifunctional structure. It has to be big enough for us to be able to analyse its behaviour without any special equipment. And last but not the least; it had to be relatively easy to get in Germany. With the help from our colleagues from biological program at the University of Tübingen, we eventually chose a very wide spread in Europe spider Tegenaria Attica.

It builds funnel webs that do not contain glue. The web can be classified into 4 main parts: shelter, (where spider spends most of his time waiting for the prey), nest (where spider lays its eggs. Normally situated inside or very close to the shelter), support­ing threads (series of structurally strong elements that support nest and shelter in space) and general network of threads (which are used for spotting the prey, as the spider itself is almost blind.

In the first week of the research we got 3 spiders: Jeff, Diane and Sparky


Which is the first thread?


As we got the spiders into our possession, we started thinking of how we analyse their behaviour. Obvious method would be documenting the web in photos and watching spider behaviour at different stages of web creation. Nevertheless those methods would not let us analyse thread orientation in the web or understand the working principles of supporting threads and how the spider lays them. Our interest at that moment could be summarised into one question. Which is the first thread?

To be able to map the step by step process we came up with a system to track spider movements. We built a box for the spider to create the web in and sur­rounded it with 3 cameras that would take a picture of whatever is happening in the box every 2 seconds. All the images are analysed by the software that we created. Software compares every next image to the original image of the empty box. It finds a changed area, which is the outline of spider body. It creates a point in the middle of the area and gives us coordi­nates of that point on the image as a text file. Using those coordinates in all 3 images and knowing the position of the cameras in space in a relationship to the box, we can triangulate 3D position of the spider at every given moment. At the moment, the software is still being developed. So far we have created a program that snaps images every 2 seconds, script that analyses and finds the positions of the points in 2d, At the moment work on transforming 2d coordinates into 3d polyline is still on-going.


Scanning the web


Second step of analysing the webs was to try to understand the geometry in detail. Capturing the web with a normal camera turned out to be a very diffi­cult task, especially on early stages of web creation, when threads are quite far from each other and the surface is not dense enough.

Using the initial tracking setup with 3 cameras, we added an infrared laser beam to the system. Moving the horizontal laser beam surface through the box with a 5 mm step we captured an image at every step. Raw output of that process is 40 images of horizontal spider web sections. Analysing them digitally and combining together in a multi-layered system allowed us to recreate general geometry of a real spider web digitally. What we did not expect though was that the scan not only gave us the gen­eral geometry, but also the thread orientation inside the web.


Key aspects


Spending some time researching the spiders we de­fined two key points of interest for us to develop fur­ther architecturally. Both aspects are closely related, they are parts of one system in different scales. On a global scale we have been looking at the concept of supporting threads in the spider web - part that structurally supports dense elements of the web. On a local scale supporting threads are very dependent on the joints between the threads. So far in fibre fabrication of the pavilions we have been using stiff materials as formwork (metal/wooden effectors), concept of supporting threads could have completely replaced that system, which would allow us to save time and money on the fabrication of the formwork. Creating a system with fibrous formwork requires a new system of fixing fibers in place. if before we used to hook carbon and glass fiber to the screws, now we would have to figure out a new system of connecting the fiber to formwork : fiber to fiber connection. We are hoping that analyzing these concepts in spi­der web can help us develop new fabrication meth­ods and tools for working with fibers.


Abstracting the joints.


Fibre to fibre interaction hasn’t been fully explored in the fibre fabrication industry yet. We have seen many examples of joints in textile and textile-based structures, but if we want to create a system with individual threads connecting with each other in a certain pattern and at specific points, there is no special tool for that.

After analysing different types of joints and its fabrication process, we started to think about how we can implement new knowledge into robotic fabrication with carbon fibre, glass fibres and resin. One of the key aspects of spiders method of fabrication is the special glue that is uses to produce joints instantly. In our case that system has to be replaced with a different method We abstracted spider’s thread to thread interaction techniques into 3 types of joints and the ways we can produce them. Our aim was to develop a tool for each type of joint that will allow us to connect two threads at any required point. First type of the joint is the “twisted joint”. In this case 2 threads are passed around each other. Second joint type is the “wrapped joint“when pone thread is passed around another one. Third type of joint is the “Intertwined joint“(pinched). Here we have 2 threads tangled around each other but not really passed around each other.­


Joint tool 01. Wrapper. (prototype)


Joint tool 01 is introducing an entirely new system of joint fabrication. It allows creating a wrapped joint Key feature of this tool is that it allows passing one thread around another without having to pass it from one robot arm to another.

The main problem that we have faced with this tool is that the bobbin with fibres has to be placed on the tool itself, which limits the geometry of the final design. Gaps in-between the supporting threads always have to be big enough to pass the bobbin in-between them. If on the other hand we use a smaller bobbin, we get less limited in geometry, but would have to replace the bobbin very often because of the short length of the thread. It is also possible to use the tool for joining the fibres with a third fibre. Than the thread that sits on the tool is only used for joints themselves and the “working“ thread spool is situated somewhere next to the robot arm, but not necessarily on it. In that case we would need to solve the problem of cutting the joint thread every time we create a joint. Also we would need to think about instant curing, otherwise joint would loosen the moment we cut the joining thread. For this we suggest to use preimpegnated fibres and UV lamp for instant joint curing installed on a tool itself. Joint tool prototype is brought to movement with a stepper motor Nemo16 which is controlled via the Arduino board.


Joint tool 02. Twister. (prototype)


Joint tool 02 works on the same principle of rotating gear with a gap inside as Joint 002. But the “Twister” creates a completely different type of joint. The tool creates a “twisted“ joint by placing 2 threads into 2 separate gaps in the rotating element and wrapping them around each other. Unlike tool 002 this one doesn’t require a bobbin placed on the tool. That allows a certain geometrical flexibility both on a design and fabrication stage. At the moment prototype is 160 mm in diameter but we believe that it i possible to create a much smaller one so that the tool can reach everywhere even in a very complex structure. This tool requires instant joint curing. Just like in the previous case we suggest using preimpregnated fibres and UV curing lamps installed on the tool itself. Joint tool prototype is brought to movement with a stepper motor Nemo16 which is controlled via the Arduino board.


Digital model 003


Digital model 003 brings model 002 into a 3D shape. Going back to the dome shape, we break it up into layers and approximate each layer using methods developed in the Digital Model 002. After creating frames that are formed by support­ing threads and the curve approximation, we fix the system by attaching the layers together. For that we use a zigzag pattern that we have seen in the spider behaviour, In that way we create an easy to control system, adapting spider web principles of supporting and overlaying threads. In the real scale production of the system (for exam­ple as a pavilion) we suggest that the tools that we have mentioned before will be used in the produc­tion process. Tool 002 is used to build the supporting threads and curve approximation and tool 003 for overlaying zigzag threads


Physical model 003


Realizing the physical model that is based on the digital one was our naturally logical next step, basi­cally we had the working drawings, of the digital model, exploded into sections. We also autotomized our method according to the following steps:

1- Building the wooden framework.

2- Installing transparent threads for external support on the surface area of the box, step size of 50 mms. 3- Installing the “Supporting threads”, fixed to the en­vironment, with 4 anchor points. They are considered the structural base of the whole model.

4- Adding the “Approximation Polyline”, the deforming lines were added on the frames with a certain tension in 10 different anchor points, in each corner with a certain step size.

5- A crucial part, was to completely cure the beam like corners, because we needed to treat the product as a whole, rather than leaving the threads to behave individually in unpredicted manner.

6- The resulting deformed sections were then stitched together with one longitudinal polyline run­ning across the whole product, giving it shape, and form.

7- The last was basically laying on the truss like sys­tem, that would ensure the product was covered and resulting in a performing shape. The laying proce­dure, was exploded into two steps, one for the upper area, and one for the bottom area.

8- After curing the whole model, we are able to de­tach the thread work out of the supporting threads.


We experimented this method on two physical model, one had the characteristics of a vault, basically a positive curvature shape. The second model, had both negative and positive curvatures. We conclude that this method is most suitable, for re­placing the heavy costly metal supportive structures, with a light quick and easily assembled thread one. Moreover, once the core of the model is fully cured, it becomes self-standing and structurally stable.