What is Emergent Behaviour?
Economies, beehives, financial markets, animal markings, team building, consciousness, locust swarms, mass hysteria, geese flocking, road networks and traffic jams, bacterial infection, town planning, evolution, the Web … these are all examples of emergent phenomena where a collection of individuals interact without central control to produce results which are not explicitly “programmed”.
Qualities of Emergent Behaviour
What can emergent systems do that other systems can’t?
- They are robust and resilient. There is no single-point of failure, so if a single unit fails, becomes lost or is stolen, the system still works.
- They are well-suited to the messy real world. Human-engineered systems may be “optimal” but often require a lot of effort to design and are fragile in the face of changing conditions. Importantly, they don’t need to have complete knowledge or understanding to achieve a goal (e.g. social systems in warehousing).
- They find a reasonable solution quickly and then optimise. In the real world, time matters – decisions need to be taken while they are still relevant. Traditional computer algorithms tend to not produce a useful result until they are complete (which may be too late, e.g. if you’re trying to avoid an oncoming obstacle) .
How it works
The individuals interact with each other directly or indirectly (via their environment). Interacting via an effect on, and response to, their common environment is called stigmergy. For example, termites work together to build termite mounds without any “queen” to co-ordinate activity and without any pre-existing plan of what to build. They change the environment and the changed environment modifies their behaviour. For example, to build a single termite mound in an environment consisting of randomly-scattered wood chips, a group of termites each has only to follow one simple rule :
Whilst wandering randomly If you find a chip then pick it up unless you're already carrying a chip in which case drop it
To begin with, several small mounds will start to emerge, but then the largest mound will grow at the expense of the smaller ones until there is only the larger one left. This is because termites are more likely to find the large mound than the small ones. You can gain an intuitive understanding of this by downloading the StarLogo application by Mitchel Resnick of MIT.
Interestingly, through emergent behaviour “selfish genes” can cause apparently social behaviour. By forming into schools (using simple emergent “flocking” rules), animals like fish and zebras reduce their individual chances of predation.
Which problems of today can emergent systems solve?
- Robotic systems capable of operating in the real world, e.g. planetary exploration, demining, domestic. Robots can share their information.
- There are a host of military applications, for example the work done by DARPA on groups of small (<5cm) distributed robots. MAVs (Micro Autonomous Flying Robots).
- Toys – a technology platform for social games?
- Financial systems, from the stock market to local and global economies, can be modelled using a “SimCity”-style simulation of thousands or millions of agents all following simple rules (e.g. “if my stock tanks then sell”). Likewise traffic flow can be modelled with agents following simple rules such as “if the car in front gets too close then brake).
Who’s working on this?
Much of the work is being done in the USA, especially at Santa Fe. Work in the UK includes:
- Intelligent Autonomous Systems Engineering Laboratory, Owen Holland and Chris Melhuish at the University of the West of England, Bristol, UK have worked on Collective Robotics.
- Cliff and Miller at Sussex and Nottingham. (but now at MIT) explored co-evolution using neural networks.
- Dr R W Taylor at York built a CA hardware engine called CAM-6
- Evolutionary and Emergent Behaviour Intelligence and Computation (EEBIC) group at the University of Birmingham
- Nic Holt of the Computer Science Research Strategy Group Forum
- Multi-agent Systems in Logic Programming
- Frank Guerin’s work on agents at Imperial College
- Cambridge ORL/AT&T‘s Piconet
Intelligent Road Stud (IRS, cats-eyes which talk to each other).
- Turing and John Von Neumann’s replicators.
- Rodney Brook’s Subsumption Architecture
- Richard Dawkins’ father created shapes that self-assembled when shaken together, (but I can’t find a description online).
- Artificial Life
Boids demonstrates flocking and other distributed behaviour.
- Self-Organising Systems
- Cellular Automata
- Cambridge University Engineering Department NanoScale Science Laboratory
Automated Robotic Design by Genetic Algorithm, which is an example of Artifical Ontogeny, also studied by Josh Bongard at the University of Zurich.
- Evolving Spiking Neurons, as used by Zufferey in his flying robot PhD project.
- Keyence Engager 4-rotor flying vehicle.
- Robot soccer tournaments.
- Videos on VEGA website:
- Coordination without Communication, a discussion by Stan Franklin at University of Memphis.
- Distributed Robot Architectures (DIRA) at Carnegie Mellon University
- Self-Assembly of 2D & 3D DNA Structures, John Reif at Duke University
- Chaos, James Gleick
- The Pattern On the Stone, Daniel Hillis
- Emergence, Steven Johnson
- Swarm Intelligence, Eric Bonabeau, Marco Dorigo, Guy
Theraulaz at Santa Fe Institute
Mambo Chicken and the Transhuman Condition, Ed
- Ashley Book of Knots, contains diagrams and descriptions of 3854 things that can be done with rope and string, virtually all of which involve some version of over and under.
- Engines of Creation,
K Eric Drexler, the quintessential nanotech promoter.
- Analog VLSI and Neural Systems, Carver Mead
- Self-Organizing Maps, Kohonen
- Brainmakers, David
- Pulsed Neural Networks, edited by Wolfgang Maass and
- A Fire upon the Deep, a novel by Vernor Vinge, an interesting insight into how distributed individuals might think.
© 2003 Pilgrim Beart