Trail Maintenance Need Prediction Proposal

Problem:

Trail maintenance is often uncoordinated and under planned. Trails, despite being many people and community’s recreation space of choice, and a vital part of maintaining a healthy wilderness by relegating traffic to specific areas, often become overgrown or unkept due to bad maintenance planning. Within a wilderness space, some trails are over frequented by maintenance groups, resulting in wasted work and time, while others are under frequented, and left to become littered with fallen trees, ruts from runoff, and overgrown shrubbery.

Our solution: 

By using machine learning algorithms to draw a correlation between slope, rainfall, traffic, soil composition, and trail maintenance needs, we intend to create a prediction algorithm for the frequency that a trail should be maintained in order to improve efficiency of trail maintenance groups and organizations. 

Our method:

Using data from multiple organizations both globally and locally that maintain trails and log work hours, and augmenting this with soil, grade, traffic, and weather data, we will create a dataset that relates external factors to trail maintenance needs. We will then implement a deep learning algorithm on this dataset, potentially additionally calculating the interactions between grade and weather (runoff), which determines necessary work hours from these external factors. Some of this data will be sectioned off from training to use for verification of the model (likely the local data). Thus, we will create a model which can predict trail maintenance needs for new or unlogged trails.