Design Strategy

The team started off by creating a prioritization model that aims to identify the neighborhoods and roads with the highest need for streetlights based on three factors of importance to the DOT: crash severity, crime rate, and streetlight coverage. Once these metrics were calculated for each neighborhood and street in Atlanta, the next step was to create a neighborhood and street ranking according to need for additional streetlights. The next component of this model was to estimate the number of additional lights needed to standardize illumination levels of roads in Atlanta. In our dark spot analysis, we determined the amount of additional lights needed to increase the percent lit to a predefined threshold, and created a map with the approximate locations of the new streetlights by going through each dark-spot on a road and progressively adding lights without buffer overlap until the defined illumination-levels were met. 

Based on our analysis of the relationship between crashes, crime, and lighting, we found that the impact of additional lighting on public safety is not the same for all neighborhoods. To more accurately assess the impact of geo-demographic factors on the possible benefit of increased lighting, we clustered the neighborhoods using the K-Means Algorithm. We based the clustering on the following attributes: percentage of local and arterial roads, population density, geographic location, and median household income. The results of the clustering allowed us to identify which neighborhoods would respond best to increased lighting.

This led us into the next part of our project: the trade-off analysis. To make recommendations, we considered both equity and efficiency. In terms of equity, it was clear that some of the clusters were in need for more lights than others, but those with characteristics such as high population density and high levels of lighting, as well as those with very low population densities and high median household income would not benefit the most from additional lighting. Our team decided to prioritize neighborhoods in the clusters where additional lighting would have the biggest impact. These neighborhoods have been typically underserved by the government, and share characteristics such as low median household income, low levels of lighting and high levels of crashes and crime. In terms of efficiency, our team performed a cost-benefit analysis for the installation of new lights in each cluster. To calculate cost, our team took into account installation, maintenance and operation for new lights over a designated timespan. The social benefit was calculated from the projected decreases in crashes and crime based on a regression model, and used societal costs for each type of crash and crime to quantify the impact.