Research & Innovation

Learning from the Swarm

Nature takes full advantages of swarm intelligence. Now IT scientists hope to take advantage of the knowledge in bird flocks and insect colonies as well.

11 Feb. 2016

German drivers spend 39 hours a year on average stuck in traffic. For residents of big cities it's even more. One big contributor is 'dumb' stoplights that run through a set program, more or less regardless of the actual flow of traffic or other conditions. Wouldn't it make more sense to develop intelligent systems that can adapt flexibly or even learn on the go?

Dr. Christian Müller-Schloer of the Leibniz University of Hannover (Hall 6, Stand A27) is researching precisely that topic: Organic Computing. "We are trying to take the mechanisms that we observe in complex natural systems and transfer them to technology," Müller-Schloer says. The self-organized navigational skills of ants offer one example. "When ants stumble upon a source of food and not long after have already discovered the shortest path back to their lair, you have to ask yourself: How can they achieve that, despite the limitations of their sensory perception?" Müller-Schloer says .


Systems adjust automatically

Organic Computing (OC) attempts to explore the foundations of this phenomenon by imitating natural systems that organize, optimize and heal themselves. The research is then used to transfer these so-called "self-x properties" into technical systems.

Traffic lights for example can be programmed to adjust intelligently to traffic. One such solution is already in planning for Darmstadt, Germany, with go-live scheduled for the middle of this year. "Organic Computing doesn't mean adopting biological principles," notes Dr. Hartmut Schmeck from the Karlsruher Institute for Technology (Hall 6, Stand A30). It instead is focused on adaptable behaviors. Intelligent traffic controls are just one example of OC.

A painful learning process

There's one catch: intelligent systems are inherently designed to make their own decisions — in some cases decisions utterly beyond what their initiators anticipated. Organic Computing has to account for this. Humans must have the final word and the ability to intervene in practically all situations. Otherwise a networked traffic light system might switch all the lights to green at once, just to find out what would happen.

For Hartmut Schmeck all roads lead to Organic Computing : "In five to ten years we'll be surrounded by a variety of intelligent systems — in our cars, in our homes, at the workplace." The systems will inherently need to communicate with one another, with or without the help of their users. "The questions isn't whether self-organizing systems will arise, it's how we'll design them."

Intelligent Systems go their own paths

One positive example might be a smartphone that mutes itself automatically in a concert hall, conveys addresses to a car's GPS system and turns on the home heating system at the right time before the owners get home. There are also potential negative examples of systems behaving in ways that might not fit our intentions, Schmeck fears.

Yet there's much more to Organic Computing than just switching traffic lights. At the Karlsruher Institute for Technology (KIT), computer scientists, electrical engineers and a neural physiologist are working on an autonomous production robot. Instead of running through a fixed set of work instructions, the machine is programmed to analyze the situation and decide what needs to be done. Autonomous robots can form ad-hoc networks to work with other robots on outstanding tasks. The swarm can detect when sub-systems are unavailable or overloaded and assign new resources to the task.

Is the breakthrough ahead?

The next generation of robots is expected to bring autonomous behavior at low costs. They must also be self-configuring, self-healing and self-optimizing. Researchers at the Leibniz University of Hannover predict that the big breakthrough in this field will come in just a few years. The key technology for this involves data processing based on biological processes — especially organic computers.

ARMAR, a humanoid robot produced by KIT, can already understand and execute spoken commands, such as fetching the milk from the fridge. It uses cameras to track its position in the room, detect objects and pick them up with the required sensitivity. It can also react to gestures, and can learn tasks such as emptying a dishwasher or wiping a counter by watching its human counterparts. In this way it adapts itself naturally to the humans around it. As such, it's only a matter of time before computer systems can organize themselves and forge their own plans on how to fulfill their mission. Hopefully for the benefit of mankind.

For deeper insights into the world of Organic Computing, visit the CeBIT Global Conferences, such as the keynote from Dr. Sami Haddadin , Director of the Institute of Automatic Control at the Leibniz University of Hannover. His lecture is entitled "Robots may not be able to think on their own — but they can feel. " (Hall 8, Open Stage, 16 March, 11:45 a.m.)

Research & Innovation Organic Computing CEBIT RSS Feed