Inspiration
We were inspired by the charging times of electric vehicles being especially long, and some drivers of EV not being able to fit an entire charging block within their schedule. We wanted to tailor a solution that would predict optimal times to charge EVs based on data analysis of the driver's driving habits and optimal battery health.
What it does
The app itself is aimed at providing a user-friendly mobile app that allows users to have a profile log in, and view the battery data and charging schedule for their car. They can select from multiple cars if they wish. The charging schedule calendar on the app takes in data from the model and outputs free blocks during which it predicts that charging would be optimal. The user receives push notifications everyday which let them know which time slots that they are free, as well as a notification when their battery is at 30% or less. If they click on the free time slot, they are led to a pop up which takes provides them with a list of electric charging stations based on proximity to their location.
How we built it
We built the model using Python, Pandas, and NumPy as we entered an initial set of mock data and trained the model to learn and perform predictions based on the data with high accuracy. For the app itself, we used Flutter to develop a user-friendly mobile app that allows users to have a profile log in, and view the battery data and charging schedule for their car. They can select from multiple cars if they wish.
Challenges we ran into
As the three of us dove into a field that was not only new to us but intimidating, we ran into several obstacles that stopped us from progressing. When training the model, an accuracy of above 90% was becoming impossible to achieve, but after trying different approaches and implementing different types of models, we were able to achieve a 100% accuracy on the predictions for the timings and battery. Another challenge we ran into was working with Flutter to integrate the model into a calendar that we could use to display the information returned from the model. We did not have enough time to use Flask or another python framework to actually integrate our model data with the calendar on Flutter.
Accomplishments that we're proud of
We implemented an AI model that works. It uses data to predict patterns within the user's schedule and provide the time slots where charging would be most optimal.
What we learned
Throughout the development of Chargeota, our team embarked on a comprehensive learning journey, gaining invaluable insights across various areas. We delved deep into data analysis and machine learning, using tools like Python, Pandas, and NumPy to understand complex user driving patterns and optimize EV charging schedules. Achieving a breakthrough from 90% to 100% accuracy in our AI model not only honed our skills in machine learning but also underscored the importance of model selection and optimization. Meanwhile, developing a user-friendly mobile application with Flutter challenged us to master UI design and app-user interaction. Integrating a sophisticated machine learning model into a mobile app framework presented unique challenges, teaching us valuable lessons in technology integration and the significance of planning and adaptability in tech development.
What's next for Chargeota
Chargeota recognizes that this isnt the end of the journey for this project. There is still much more to do including updating the UI and implementing more user features. Utilizing Flutter's diverse tools, we will continue to rework the user experience until there are no hiccups in the presentation of the model. Additionally, we can also continuously improve the model to achieve higher quality data. We can perform more specific predictions and can have it provide suggestions on how to improve driving to optimize battery performance, tailored to each user.
Log in or sign up for Devpost to join the conversation.