Steve Bono

10/31/2002

Artificial Intelligence

Fuzzy Traffic Lights

(Topic Selection)

I propose to create a system for the operation of intelligent traffic lights. Conventionally, traffic lights are time-based or sensor-based, both of which function far below optimal performance. My proposed system will take into consideration the various goals and purposes for having traffic lights, and use a fuzzy logic control unit to automate the changing of the lights to produce the best results.

System Goals

1. Uniformly distribute the average waiting times for each car at the intersection.

2. Maximize the throughput of the intersection.

Why is this system important?

Conventional traffic lights cycle on static timers, cycle on semi-dynamic timers that change cycle lengths throughout the day, cycle based on sensory input such as motion or weight sensors, or cycle at the hands of a human operator. All of these methods have setbacks that can be accounted for with an automated fuzzy traffic light system.

It should be obvious that lights operating on timers are the worst case. They have no way of determining whether or not any cars are even at the traffic lights. Sensory traffic lights fail because during heavy traffic they cannot predict the directions of traffic that are in greater need of a longer traffic light cycle. A human operator comes as close as possible to achieving the best case scenario out of the existing systems, however, I believe that given the large number of scenarios resulting from complex intersections, a fuzzy logic system (though more difficult to construct than hiring a person) will operate far more effectively than a human operator, and once developed will be reusable in many different situations.

Current and previous research on this topic

Since I have started researching this topic I have only found a few sources directly relating to the subject of fuzzy logic traffic lights. Though the few I have found are very detailed and informative. There are however a large number of fuzzy logic resources to aid in my research. I do not plan to mimic or even follow in the least any implementations of an intelligent traffic light that I find. I plan to use them as resources for evaluating my system and possibly filling in any holes I might have.

I can provide you with a list of the resources I found so far ahead of time if necessary.

System Requirements

(Note: These sensors are assumed to exist and their design is not taken into consideration for this project)

1. The system must take input from various sensors to have access to the following information:

The Number of cars at the intersection in each direction. (Necessary)

The speed of the cars going through the intersection. (Ideal but not necessary)

2. The system must have a timer.

Proposed plan for research and development

Initially, I plan to only consider the case of a 2-way intersection (possibly 4-way), since they seem to be the easiest models.

1. I will design a set of needed inputs and outputs for the system to operate optimally and then design the system.

2. I will then create a simulation, most likely using C++ since it is the language I am most familiar with.

3. I will analyze the results of my simulations and report on them.

4. Once I have done this for a typical 2-way intersection, I will extend it to account for various other more complicated intersections (Intersections containing multiple directional arrows, etc…).

Within the time given for this project I think I will be able to complete a basic model and do fairly extensive analysis of the model, and if time permits I will consider the more complicated situations.