Swarm Paintings: Non-Human Art

General Background: Since childhood i have enjoyed sketching and painting, and very much at that! Sometimes i found myself copying an existing image or painting, making small changes here and there in it. Yes, the paintings came out beautiful (or so i think!), but one thing always made me unhappy, i thought that the creativity needed to make original stuff was missing at times (not always). It was not there all the time. It came in bursts and went away.

I agree with Leonel Moura (from his article) that creativity is basically produced due to different experiences and interactions. Absence or lack of which could make art lose novelty.

Talking of novelty, how about looking at art in nature? Richard Dawkins states that the difference between human art or design and the amazingly “ingenious” forms that we encounter in nature, is due tho the fact that Human art originates in the mind , while the natural designs result from natural selection. Which is very true. However it is another matter that natural selection and cultural selection, that will ultimately decide on the “popularity” of an art don’t function in the same way. Anyhow How can we remove the cultural bias or the human bias that we have in our art forms?

.

Answers in Artificial Life: Artificial life may be defined as “A field of study devoted to understanding life by attempting to derive general theories underlying biological phenomena, and recreating these dynamics in other physical media – such as computers – making them accessible to new kinds of experimental manipulation and testing. This scientific research links biology and computer science.”

Most of the A-Life simulations today can not be considered truly alive, as they still can not show some properties of truly alive systems and also that they have considerable human bias in design. However there are two views that have existed on the whole idea of Artificial Life and the extent it can go.

Weak A-Life is the idea that the “living process” can not be achieved beyond a chemical domain. Weak A-life researchers concentrate on simulating life processes with an underlying aim to understand the biological processes.

Strong A-Life is exactly the reverse. John Von Neumann once remarked “life is a process which can be abstracted away from any particular medium“. In recent times Ecologist Tom Ray declared that his computer simulation Tierra was not a simulation of life but a synthesis of life. In Tierra, computer programmes compete for CPU time and access to the main memory. These programs are also evolvable, can replicate, mutate and recombine.

.

Relating A-Life to Art: While researching on these ideas and the fact that these could be used to generate the art forms that i talked about in the first paragraph i came across a few papers by Swarm Intelligence Guru Vitorino Ramos and a couple of articles by Leonel Moura who had worked in collaboration with Dr Ramos on precisly this theme.

.

Swarm Paintings: Thus the idea as i had mentioned in my very first paragraph is to create an organism ideally with minimum pre-commitment to any representational art scheme or human style or taste. Sounds simple but is not so simple to implement!

There are a number of projects that have dealt with creating art, but these mostly have been evolutionary algorithms that learn from human behavior, and learn about human mannerisms and try to create art according to that. The idea here is to create art with a minimum of human intervention.

I came across a project by Dr Vitorino Ramos to which i had pointed out implicitly in the last paragraph. This project called ARTSBOT (ARTistic Swarm roBOTs) project. This project tries to address this issue of minimizing the human intervention in aesthetics , ethnicity, taste,style etc. In short their idea was to remove or to minimize the anthropocentric bias that pervades all our art forms. Obviously all this can have massive implications in our understanding of the biological processes also, however here we’ll talk of only art.

Two of the first paintings that emerged were:

(Source: Here)

(Source: Here)

.

These paintings were among the first swarm paintings by Leonel Moura and Vitorino Ramos. Now we see that these seem detached from a functional human pre-commitment. They don’t seem to represent any emotion or style or taste. However they still look very pleasant!

However the point to be understood and to be noted is that these are NOT random pictures created either by a programme or by a swarm of robots moving “randomly”. These pictures were generated by a horde of artificial ants and also by robots. They are not random, but they EMERGE from a process of pheromone deposition and evaporation as was simulated in this system from ants. Thus the result that we have above is a Colony Cognitive Map. The colony cognitive map is analogous to a cognitive map in the brain. I will cover the idea of a colony cognitive map in the next post.

A couple of more beautiful paintings can be seen below!

(Source for both images : Here>>)

Though i have already mentioned how these art forms emerge, i would still like to quote a paragraph from here:

The painting robots are artificial ‘organisms’ able to create their own art forms. They are equipped with environmental awareness and a small brain that runs algorithms based on simple rules. The resulting paintings are not predetermined, emerging rather from the combined effects of randomness and stigmergy, that is, indirect communication trough the environment.
Although the robots are autonomous they depend on a symbiotic relationship with human partners Not only in terms of starting and ending the procedure, but also and more deeply in the fact that the final configuration of each painting is the result of a certain gestalt fired in the brain of the human viewer. Therefore what we can consider ‘art’ here, is the result of multiple agents, some human, some artificial, immerged in a chaotic process where no one is in control and whose output is impossible to determine.
Hence, a ‘new kind of art’ represents the introduction of the complexity paradigm in the cultural and artistic realm.’

A Painting bot is something like in the picture shown below:

A swarm of robots at work:

The final art generated by the swarm of these robots is beautiful!

(Photo Credit for the three pictures above: Here>>

.

Conclusions:

The work of Dr Ramos and Leonel Maura can be summed up as:

1. The human is only the “art-architect”, the “swarm” is the artist.

.

2. The “life” of Artificial Life shows characteristics like natural life itself namely Morphogenesis, ability to adapt to changing environments, evolution etc.

.

Leonel Moura’s wonderful article states that the final aim is to create an “Artificial Autopoietic System”, intriguing indeed and eagerly awaited!!
Such simulations could change the way we understand the biological processes and life.

Also i am now thinking how could music be produced based on the same or similar ideas. I wonder if Swarm music could be available. It would be most interesting and i can’t wait to listen to it already!

.

Have a look at this video by Leonel Moura, having some time lapse footage of robots painting.

References:

1. Ant- Swarm Morphogenese By Leonel Moura

.

2. On the Implicit and on the Artificial – Morphogenesis and Emergent Aesthetics in Autonomous Collective Systems, in ARCHITOPIA Book, Art, Architecture and Science, INSTITUT D’ART CONTEMPORAIN, J.L. Maubant et al. (Eds.), pp. 25-57, Chapter 2, Vitorino Ramos.

.

3. A Strange Metamorphosis [From Kafka to Red Ant], Vitorino Ramos

.

Links:

Follow the following links to follow on more exciting papers and paintings.

1. Dr Vitorino Ramos

2. Leonel Moura

Onionesque Reality Home >>

Advertisement

Read Full Post »

Adaptive Routing taking Cues from Stigmergy in Ants

There are four distinct advantages in using self organized, emergent, decentralized or “swarm intelligent” systems ^[1], these are in fact the benefits of a social insect colony as well. We see these colonies as the inspiration to developing artificial decentralized systems.

1. Flexible: The system can react to internal and external changes.
2. Robust: Tasks are completed even if some individuals fail
3. Decentralized: There is no central control over the system
4. Self Organized: Paths to solutions are emergent rather than pre-defined which leads to better utilization of resources and gives a dynamic character to the solution finding process.

Ants Behavior, Stigmergy and cues in Routing
Ants are very simple insects but collectively they can perform complex tasks with good consistency. Examples of such complex problem solving behavior have been documented, and a few examples are as follows.

1. Building nests (ant hills) and maintenance.
2. Regulating nest ambient temperature fluctuations very low and maintaining required levels of oxygen with small variations.
3. Co-operating in carrying large prey.
4. Finding the shortest routes from the nest to the food.
5. Prefentially exploiting the richest food source available.

Stigmergy refers to communication indirectly, by using markers such as pheromones in ants. In the above examples, two distinct types of stigmergy are observed. One is called sematectonic stigmergy, it involves a change in the physical environment characteristics. An example of sematectonic stigmergy is nest building wherein an ant observes a structure developing and adds its ball of mud to the top of it. Another form of stigmergy is sign-based and hence indirect. Here something is deposited in the environment that makes no direct contribution to the task being undertaken but is used to influence the subsequent behaviour that is task related.

Here is an example of what sematectonic stigmergy can achieve (one has to keep in mind that individual termites are very simple creatures):

Sign based stigmergy is very highly developed in ants. Ants use chemicals called as pheromones to develop a very sophisticated signaling system. Ants foraging for food lay down some pheromone which marks the path that they follow. An isolated ant moves at random but an ant encountering a previously laid trail will detect it and decide to follow it with a high probability and thereby reinforce it with a further quantity of pheromone. Since the pheromone will evaporate the lesser used paths will gradually vanish. We see that this is a collective behaviour.

The foraging behavior in ants can be best understood with the following illustrations.

Two ants start with equal probabilities of going on either path.

Ant taking the “lower” path has a shorter to and fro time from the nest to the food.

The pheromone density on the path traversed by ant on the lower path is greater because of the two passes by that ant on this path. Hence other ants say C and D are bound to follow this path because of the stronger scent.

Over many iterations many ants start using this path thus further reinforcing it and after a while this path i.e the shorter path is almost exclusively used.

So we see how randomness is followed by “positive feedback” which causes amplification and the consequent exclusive use of the shortest path. This is an apt example of how ants choose shortest paths.

Now if the shorter path is blocked or unavailable, then in this scenario the longer route may still be used and made the preferred route by repeated use(positive feedback) and the consequent amplification.

By repeated use of the other path, the pheromone concentration here gets stronger that leads to that path being used. Thus this is how ants develop a solution when a path is blocked. Thus this illustrates how swarm intelligent systems adapt with changes in the environment.

Mathematically we put the above as : Suppose the distance between two points i and j is d_ij and τ_ij is the pheromone concentration on the link ij. Suppose ‘m’ agents are traversing this link and building a tour. At each step of the tour, the probability to go from one point ‘i’ to another point ‘j’ is (τ_ij)^a (d_ij)^-b .

After building a tour of length L each agent reinforces the edges it has used by an amount proportional to 1/L. Also the pheromone evaporates such that τà(1-ρ)τ.

The above is used to solve the traveling salesman problem as well. There are a few key concepts that we gather from the above illustrations.

A) Positive feedback – build a solution using local solutions, by keeping good solutions in memory.

B) Negative feedback – want to avoid premature convergence, evaporate the pheromone.

C) Time scale – number of runs are also critical.

These key points then in turn can be summed up for an ant colony as very simple rules: lay pheromone, and follow trails of other ants.

Applications to Routing Scenarios
Routing is the mechanism that handles and directs messages at switching stations. The important points we see are that messages must reach their destinations, it should take as little time as possible to reach from source to destination, and also the traffic is constantly changing hence routing should adapt accordingly. The inspiration for using the example of ants for routing in communications networks arises from the fact that the present routing systems depend upon global information for their efficient operation. Ant systems on the other hand, rely on pheromone traces that are laid down in the network as the ant moves through the network. Global information is frequently out of date and transmission of the information required from one node to all others consumes considerable network bandwidth. Ideally, we would like to have the network adapt routing patterns to take advantage of free resources and move existing traffic if possible.

Now if we consider a telephone network. A telecommunication environment is highly unpredictable. With delays or problems when least expected at times. Suppose a phone call is made form Mumbai to Shanghai. This phone call has to go through several intermediate steps/nodes such as maybe Kolkatta and Bangkok etc. such a system requires a routing mechanism that will tell the call where it should hop next to establish a connection, a good routing technique as we have already listed would minimize this time (ants finding shortest path) and avoid congestions. Backup routes are very important in such a system. In some special circumstances when there is an explosion or if there is a news programme that requires calling, then it leads to localized surges in the phone traffic. This then requires the phone traffic to be re routed to the less congested parts of the network (ants adapting and finding an alternative path).

This can be simulated as follows. Hordes of agents can roam the network and leave bits of information which can be thought of as an analogous to pheromone to reinforce paths through uncongested areas. Phone calls are then routed by following the trails of these ants like agents. To exactly mimic the ants working, these “pheromones” can be evaporated by a certain mechanism by a rate that may be decided by certain considerations. Now if suppose a certain path that was very swift becomes congested then the agents to this route can be delayed by certain time spans. This time span will allow the virtual pheromone to evaporate and thus reinforcement process can be overcome. Hence after a while this route is abandoned. The ant-like agents (as seen above) can discover new paths to re-route the traffic. The advantages of such a system are: the calls get forwarded at a good speed and also the congested areas gradually get de congested. Also the bucket brigade process makes the routing process more robust.

A detailed approach to ant based routing has been described by White (1996)^[2]. Such systems have three types of agents: explorers, allocators and deallocators. Explorer agents mimic the foraging behaviour of ants and follow trails of pheromones laid down by previous explorers (as quantitatively described in the Mumbai- Shanghai phone call example). Allocator agents move across the paths determined by explorer agents and allocate the bandwidth on the links used in the path. Similarly, when the path is no longer required deallocator agents traverse the path and deallocate the bandwidth used on the links.
References:
[1] E. Bonabeau, M. Dorigo, and G. Theraulaz, Swarm Intelligence: From Natural to Artifcial Systems. NY: Oxford Univ. Press, 1999.

[2] White A.R.P., Routing with Ants, Nortel internal report, 1996.

Read Full Post »

The Working of a Bird Swarm

This post is following the previous post in which i mentioned about the rules governing the motion of a bird swarm.

The video below is a breathtaking, sublime, amazing recording of thousands of starlings in a flock before roosting. This is from the UK country-side. While it is another awesome demonstration of how the “imagination” of nature can be like, it also gives a perfect example of how the entire swarm is organized and how it moves. Modern researchers are trying to imitate such emergent behavior to use in fields like robotics, data mining, internet mathematics, optimization etc etc.

How do birds exactly do this?

Here is an excerpt from a talk that i had given at a technical event in October ’07.

In a swarm if we say that there are ‘N’ number of agents, then we can say that these autonomous agents are in a way co-operating to achieve a global objective. This global objective can be better foraging, constructing shelter, serving as a defence mechanism among others. This apparent collective intelligence emerges from very simple individual agents. The actions of these agents are governed by local rules and through the interactions of the N agents the swarm achieves a global objective. A kind of “self organization” emerges in these systems. We see that there is no central controller in such cases. Swarm intelligence gives a basis which makes it possible to explore collective (or distributed) problem solving without centralized control or without the provision of a global model. ^[1]

The individual (but autonomous) agent does not follow directives from a central authority or work according to some global plan. As a common example, a bird in a flock, only adjusts its movements to coordinate with the movements of its flock mates or more precisely the members that are its neighbors. It simply tries to stay close to its neighbors, but avoid collisions with them. Each bird does not take commands from any leader bird since there is no lead bird. Any bird can fly anywhere in the swarm, either in the middle or the front or the back of the swarm. Swarm behavior gives the birds some distinct advantages like protection from predators, and searching for food (effectively in a swarm each bird is exploiting the eyes of every other bird). Scientists are trying to find out how these birds, fish etc move in flocks, schools in a way that appears orchestrated. A school of fish and a flock of birds move as if all the “steps” were pre planned. For one moment they are moving towards the left and in another they are darting towards the right. Among these researchers in 1987, Reynolds created a boidor bird-oid (bird like) model. This is a distributed behavioral model, to simulate the motion of a flock of birds on a computer ^[2]. Each boid is implemented as an agent which moves according to its own understanding of the dynamic environment. A boid observes the following rules. First, is that a boid must move away from boids that are too close, so as to reduce the chance of collisions. Second, boid must fly in the general direction that the flock is moving. Third, a boid should minimize exposure to the flock’s exterior by moving toward the perceived center of the flock. Flake later ^[3] added a Fourth rule, a boid should move laterally away from any boid that blocks its view. This boid model seems reasonable if we consider it from another point of view, that of it acting according to attraction and repulsion between neighbors in a flock. The repulsion relationship results in the avoidance of collisions and attraction makes the flock keep shape, i.e., copying movements of neighbors can be seen as a kind of attraction.

This is what i was talking about in the previous post. A certain set of rules is followed that would give a certain shape to the swarm, if the set is altered so will be the shape and maybe functionality as well!

Sounds nice, now how can these simple rules be modeled on a computer? They can be done using NetLogo. Here is a sample model . To get a good idea about the intricacies adjust the population, the turn angles and the vision. For playing around with the model and making your own you are instructed to go through the writeup to this model on the above page.

References:

[1] E. Bonabeau, M. Dorigo, and G. Theraulaz, Swarm Intelligence: From Natural to Artifcial Systems. NY: Oxford Univ. Press, 1999.

[2] C. Reynolds, ‘Flocks, herds, and schools: A distributed behavioral model,” Comp. Graph, vol. 21, no. 4, pp. 25{34, 1987.

[3] G. Flake, The Computational Beauty of Nature. Cambridge, MA: MIT Press, 1999.

Read Full Post »

Morphogenesis and Swarm Robotics

Here is an amazing video!

Something about this video first from the you-tube information about it.

An introduction to swarm intelligence, swarm robotics and morphogenesis. This video won the Best Video award at the AAAI-07 conference Vancouver Canada. The scientific research was performed by Anders Lyhne Christsensen, Rehan O’Grady and Marco Dorigo. The video was directed by Andreia Onofre.

Swarm Intelligence is an AI technique that is based on the collective behavior of decentralised and self organised systems. Take an ant colony for example, an ant hill is an extremely complex structure, with even things like temperature control in place. It is difficult to imagine a simple ant doing any of it! So it is basically the “swarm” of ants that is responsible for such emergent and intelligent solutions though the “constituents” of the systems (i.e ants) are simple and independent “units”. Like ants also can pick up prey many times their weight by forming very precise structures encircling the object under consideration.

like in the picture below the ants take up a comparatively much heavier fly.

photo credit with full regards : here>>

The video basically explains the same and also gives an idea on how it could be done using robots. For example a bird swarm can make very different and complex shapes depending on the set of rules under use (which in turn will depend on the scenario). I will accompany this statement by a post on a bird swarm and how they do it sometime soon. If you change the set of rules you could change the shape you get, and most of these shapes could be used to perform some intelligent task. This gives the swarm a lot of flexibility. This is basically what is Morphogenesis.

Morphogenesis could be used by a swarm of robots to move heavy objects (as an illustration to this have a look at this video, in which a group of robots pull away a child. The video can be seen here>>) which could be used in fire-fighting applications etc, to bridge gaps etc. A swarm of nano-bots could use morphogenesis to perform very specific tasks inside the human body!

Please have a look at this award winning video!

I personally thank and congratulate the director of this video for putting it all so succinctly in a matter of under 5 minutes.

Read Full Post »