Problem statement

"50 years of parking woes" -Retriever weekly

UMBC has a major parking problem. There are 7250 parking spots and 11,276 issued permits. That means an oversubscription by 55%. To put things into perspective, UMBC students, faculty and general staff spend 76 years annually looking for parking. In those 76 years, they burn 178,000 gallons of fuel worth $76,000.

State-of-the-Art (Current research)

The two popular techniques in use currently to detect open parking spots are Sensors and Computer Vision.

Sensors : Although sensors have a very high accuracy rate, they are very expensive to deploy and run. By my estimate, UMBC will have to shell out $10 million to put sensors in the parking lots.

Computer Vision : A camera takes a picture of the parking lot every minute and sends it to a central server, where the system using sophisticated techniques, determines the positions of empty parkings. The downside with this approach is that it suffers from very low accuracy. The main reasons are how the parking is laid out around campus and occlusion from trees and buildings.

Park Pronto approach

Park pronto utilizes a crowdsourcing approach to solve this problem. The users can see the available parking spots, categorized by permit types, on a mobile app. They can then also leave feedback about the current occupancy of a parking lot.

But, that's not all. Park Pronto is not just another dumb crowdsourced parking aggregator. Firstly, it maintains a 'naughty' & 'nice' score of all it's users, based on their feedbacks. It also induces a periodicity function from past data. It then reinforces the system by using the user's realtime updates. The key features are lot id, time of day, occupancy, user score and day of week. Using these features, Park Pronto is better able to gauge parking occupancy levels when faced with insufficient or untrustworthy user recommendations.

Accomplishments that I'm proud of

Park Pronto tries to solve a very burning issue in UMBC. Parking woes. It gives a new spin on an existing problem which currently has no good solution.

What's next for Park Pronto

In the coming weeks, we will be launching the app on both Android and iOS. In the future, we would also integrate with Apple CarPlay and Android Auto for more hassle free integration. The learning algorithm will get better at predicting parking lot occupancy levels over time from the information and patterns it learns from previous data.

Share this project:

Updates