SafeNav

Inspiration

SafeNav was inspired by the need to address a critical gap in navigation for people living in underdeveloped or high-risk areas. While most navigation apps prioritize speed, we recognized that safety is equally important, especially in regions where road conditions and accident rates make travel more dangerous. Communities in these areas often lack access to tools that can help them avoid hazardous routes. Our mission with SafeNav is to create a more inclusive navigation experience, offering everyone the ability to travel safely, no matter where they live or where they’re going.

What it does

SafeNav provides users with two route options: the quickest route and the safest route. Using machine learning algorithms (ex. random forrest) trained on crash data from over 200,000 incidents in Virginia, SafeNav assigns a safety rating to each crash location and calculates optimized routes that factor in both speed and safety. The app then presents these routes on an interactive map, allowing users to choose the one that best fits their needs. Whether prioritizing safety or speed, SafeNav empowers users to make informed travel decisions tailored to their comfort and security.

How we built it

Training the model: To begin with the Virginia crash data was thoroughly cleaned, as we removed over half of the features along with 800,000 rows (leaving just 2023-2024). We then took a 1000 random arbitrary data points and evaluated our own safety metric rating for the road from 1-100 for every point to use as a baseline for training the algorithm. We then trained and tested the algorithm using Random Forrest which is notorious for its handling of both classification and regression problems, and applying the results to the rest of our data points.

Integrating Google Maps: Next we constructed a script using flask (to integrate with the javascript frontend interface) that took the safety values along with the many different routes and calculated a safety score for each route, which we used along with the length in order to create our final recommendation.

Implementing Maps: We used SSL to secure encryption between the flask server and our own hosted javascript server, then integrated the program into our map on our front end system to run every-time an origin and destination is entered.

Website Design: Our website was design based off of HTML, CSS and JavaScript. Our team wanted to have a modern looking website with a focus on making the website easy to navigate for users. To do so, we made the format of the page simple in its looks with unique animations throughout.

Challenges we ran into

Given such a large dataset, both cleaning the data, dealing with outliers and, figuring out how we would set our own arbitrary data points were all problems dealing with data analysis. Additionally the sheer size of the data set drastically affected our learning strategies as we were forced to use a simpler model (random forrest instead). However, the largest challenge we ran into was our lack of Front-End to Back-End experience as connecting the flask server hosted on our virtual environment with the javascript http server caused us countless issues which we used SSL to solve. It was only through discussion and teamwork were we able to overcome all our problems and succeed.

Accomplishments that we're proud of

We're most proud of the idea behind SafeNav. Not only do we hope that SafeNav can directly help saves lives, but we also believe that the thought behind SafeNav can inspire others to build similar apps. The fact that we were able to plan and execute such a project in just 2 days shows how a group of 4 people's determination can go so far.

What we learned

Throughout the creation of SafeNav we learned about the history of car crashes and what their impacts are. On the technical side, we learned more about the usages of APIs, AI, and machine learning models. Most importantly, we learned about the importance of collaboration and dividing work to successfully complete this project.

What's next for SafeNav

In the future, we see the possibility for SafeNav to become more accurate and widespread. That involves further improvement to our ML model through better cleaning and more robust training with more applications. Additionally, we want to grow SafeNav nationwide, not just in Virginia. The larger SafeNav gets, the more lives that can be saved.