Self Organizing Systems –

Autonomous Networks, Traffic Control and Intelligent Agents

Brit Cruise

Autonomous, self-organizing, adaptive systems

A popular question in physics, chemistry and biology is “where does order come from?”. The general laws of thermodynamics tell us that systems will follow the path of least resistance to dissipate any energy they contain. Eventually the system finds its lowest energy state and will remain in equilibrium until acted upon by an outer force.

Yet in nature we can observe many systems that maintain a high internal energy and organization which seems to defy the general laws of physics. An ant grows from a singled celled zygote into a complex multi cellular organism, and then participates in a highly structured hive society. What is so fascinating is that the organization seems to emerge spontaneously from disordered conditions. The laws that may govern this self-organizing behavior are not well understood, if they exist at all. It is clear that this non-linear process is based on positive and negative feedback loops among components at the lowest level.

The study of self organizing systems is all based on a “bottom up” approach. For example and natural “hive” for “flock” system is based on each individual’s local assessment of the general conditions. The principal of locality means that every event or interaction has both local and long range effect. For example, when a tree calls in the tropics, it often drags down several other trees and leaves a gap in the canopy. This results in ecological changes, seed distribution changes and relative soil abundance. Ignoring locality obscures many factors that may alter the entire systems dynamics.

A self organized system is one which is not entirely directed by top-down rules, although their may be global constraints on the system. The local interaction of individual agents generates ordered structures at higher levels with recognizable dynamics. Since these systems are based on the subtle differences between the individual component interactions, the dynamics cannot be understood by breaking the system down into is constituent parts. This means the study of SOS must be done synthetically rather then analytically. Several research institutes are involved in this topic, such as the Santa Fe Institute (Bacon) and the Center for Complex Systems Research (Foster).

Applications of SOSThe applications of SOS are endless and are constantly being applied to new mathematical, social and physical problems. We plan to apply this theory to an autonomous traffic control system. The usability and effectiveness of traffic management and control systems depend greatly on their ability to react to changing conditions and patterns. At the Delft University of Technology researchers are proposing a system that autonomously can adapt itself in changing environments. Their UTC (Urban Traffic Control) model is primarily based on several intelligent Intersection Traffic Signaling Agents (ITSA) and some authority agents. In this way they get an UTC system based on agent technology that can adapt and respond to traffic conditions in real-time. This will allow better use of intersection capacity while maintaining stability within the overall transport system. Signal control systems have the ability to optimize and control traffic light settings in a changing environment.Specifically, in our project we plan to design our system of traffic lights with a signal control system. Each traffic light will act as a single agent, which is interconnected with all other traffic light agents. There are several current methods in which to interconnect a series of agents in such a system. More importantly, the interactions, decisions and algorithms which control and optimize the global system performance are largely unknown in science today. Researchers at DUT are still engaging in further research; “The sole optimization of one junction could induce several problems for others, so a global optimization process / optimization control via authority agents should be researched. Relatively little is known about this subject.”(Roozemond). Using computer software such as Netlogo traffic simulations can be visualized and measured to mimic actual traffic conditions (Wilensky). This software will allow us to measure the success of different control algorithms before we implement them into our physical designs. In conversations with Dr.McIsaac from the

University of
Western Ontario, the idea of using genetic algorithms to solve the optimization problem was discussed (McIsaac). If a vector of variables could be isolated to describe the behavior of both the traffic lights and autonomous cars, then an evolutionary simulation could find the solution. However the idea of using GA’s on a adaptable system which is in constant change radically new idea (Heitkötter). Little research can be found on such a simulation, so it will be approached experimentally with different software combinations. Using both a Netlogo simulation in parallel with a genetic algorithm application could be a possible solution. Once a set of behavior variables is found, it can then be simulated and implemented in our physical environment.

Limitations of the Current TechniquesTraffic lights generally employ a similar strategy, if one at all. The simplest lights run through a basic timing cycle, x seconds green then y seconds red. More sophisticated lights change their timings depending on the time of day. And the most advanced lights can change their timings depending on input from a sense. A more advanced light does not translate into better traffic control however; the overall effectiveness of most lights depends on their neighbors. Any given light, regardless of how smart it may be, must remain synchronized with adjacent lights to remain effective. For example, a light may detect that a left turn lane is empty and attempt to change its timing to take advantage of this fact. But it is still held to a static schedule, hence it cannot make full use of any special knowledge.The MONITOR system in
Milwaukeeis already part of the way to a full AI system (Doshi). They have sensors throughout the city in the form of counters, cameras, and pressure sensors. The system can detect accidents and reroute traffic, all controlled from a central location. The main problem is that its still human controlled. But the first step has been taken; those sensors are a gold mine of information. A company called UIC is already using the information to develop models. Once a system is developed that has the capability to manage the traffic, it will be relatively easy to implement it.

Autonomous CarsThe future of autonomous (auto piloted) cars will not be a system of cars driving in random directions, as our system will seem to be. With satellite navigation technology already popular in most newly manufactured cars, human drivers are finding them selves following the directions of the onboard navigation system. If we eliminate this ‘middle man’ procedure and let the cars computer decide how to get to a final destination, we have an autonomous car. The use of a system of cars in our simulation is a tool to prove how the interactions of multiple autonomous systems can interact together safely and more efficiently then ever. Current research involving autonomous cars has been successful, yet unattractive to consumers and car manufacturers. The first fleet of autonomous cars emerged from the Automated Highway System (Congress) program funded by
America’s transport department in the 1990s. After spending $14 million, AHS researchers showed off their work in 1997. They implanted 92,000 guidance magnets along a closed seven-mile stretch of motorway near
San Diego, California and then released more than a dozen vehicles on to the road. But while the cars performed flawlessly, the program was cancelled in 1998.
The development of autonomous vehicles has not ceased. In some respects, the technology is sneaking into cars already. Carmakers are starting to incorporate radar-based sensors, micro-cameras and intelligent cruise-control systems all of which were once limited to research vehicles or their luxury models. As such equipment becomes more widespread, it could make retrofitting cars for autonomous operation far cheaper and easier. For the trucking industry, an industry with more than 100% turnover rate and huge safety concerns, autonomous transportation could have an important positive impact. Already, Mercedes has developed some vision based warning systems for use in trucks (Birch).Karl Hedrick, is an engineer at the

University of
Californiaat Berkeley who spent several years developing AHS technologies in the mid 1990s and points out that it is a wary public, rather than inadequate technology, that is hindering the widespread introduction of autonomous vehicles. Dr. Hedrick suggests that that a more realistic introduction of autonomous vehicles could be designated to specialized lanes separated by fixed barriers (Behar). As drivers become comfortable with the technology they may consider converting there own vehicles.
Our autonomous system of vehicles and traffic control could easily be first applied to smaller transit systems in order to introduce its abilities. Rail lines, bus systems, corporate transportation robots are all examples of environments which would be the first to adapt an autonomous control system long before we consider converting our major highways. An extension of our research would involve the communication between adjacent cars feeding speed, traffic and weather conditions back and fourth. We are limited to only basic local proximity assessments due to our time, material and budget constraints. However, our simulation will still provide great insight into the potential of these technologies.

References

1. Bacon, D. (n.d.). Retrieved Feb. 03, 2005, from Dynamic social systems Web site: http://www.santafe.edu. 2. Foster, Glenn. Reasearch . Center for Complex Systems Research . 01 Feb. 2005 <http://www.ccsr.uiuc.edu/web/Research.html&gt;.3. Roozemond, Danko. “Urban Traffic Control.”

Delft
University. 01 Feb. 2005 <http://citeseer.ist.psu.edu/cachedpage/280175/2&gt;.
4. Wilensky, U. (1999). NetLogo. http://ccl.northwestern.edu/netlogo. Center for Connected Learning and Computer-Based Modeling. Northwestern University,
Evanston, IL.
5. Doshi, Prashant. “Automating the Evolution of Linguistic Competence in Artificial Agents.” UIC Artificial Intelligence Lab. 01 Feb. 2005 <http://enterprise.ai.uic.edu/&gt;.6. Heitkötter, Jörg . ENCORE, the EvolutioNary COmputation REpository network . 03 Feb. 2005 <http://www.cs.bham.ac.uk/Mirrors/ftp.de.uu.net/EC/clife/&gt;.7. Congress , Nita. Automated Highway System. 29 Jan. 2005 <http://www.tfhrc.gov/pubrds/summer94/p94su1.htm&gt;.8. Birch, Stuart . Automotive Engineering. 25 Jan. 2005 <http://www.sae.org/automag/electronics/08-2002/page2.htm&gt;.9. Behar, Michael. “Drivers Wanted.” The Economist. 28 Jan. 2005 <http://www.michaelbehar.com/economist/driverswanted_03_11.html&gt;.10.McIsaac, Ken. Multi-robot coordination. 04 Feb. 2005 <http://www.eng.uwo.ca/people/kmcisaac/&gt;.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: