Stoplight Optimizer

This is a program written to optimize the performance of a 3x3 grid of stoplights using a neural network.  The network attempts to maximize vehicle throughput based on a variety of start-points for the vehicles.  Next iterations of the project will attempt to implement "mass-hole" drivers, or drivers that are fairly lead-footed.

Development of the Grid

The GUI for the grid is written in vPython, containing classes for roads, cars, stoplights, and intersections.  Stoplights are able to communicate with one another in order to optimize the grid "on-the-fly" if necessary.

Results

Manually programming stoplights and relying on estimations of vehicle throughput for each intersection leads to a complicated problem with no good solution.  While one-way roads (in dense areas) and roundabouts (in more rural areas) can significantly alleviate some of these issues, these can only help so much.

Once the network is optimized, we observe a change of XX% in vehicle throughput from randomness, as well as a XX% increase from manually tuning the network.