PocketParker: Pocketsourcing Parking Lot Availability (Complete)
Parking lots present a difficult search problem. Drivers lack the visibility to determine where spots are available, and may spend a non-trivial amount of time searching for a spot. Searching not only generates frustration but also wastes energy and produces harmful carbon emissions. And while several research projects have previously attempted to solve this problem, their solutions include requirements rendering them impractical, such as additional infrastructure, on-vehicle equipment or vehicular networking, or onerous manual user input. In contrast, our solution, which we call PocketParker, requires no additional infrastructure, no vehicle modifications, and no user input, only installation on a small percentage of the millions of smartphones already in use.
PocketParker runs unattended in the background and uses the accelerometer to detect parking lot arrivals and departures based on transitions between walking and driving (departure) and walking and driving (arrival). Events are forwarded to a central server where they are incorporated into per-lot availability models allowing PocketParker to accurately order lots by the probability that they contain an available spot. In general, we consider our approach to be an example of a subset of crowdsourcing that does not require any manual user input, which we call pocketsourcing.
We have designed, built, and deployed at PocketParker prototype at UB. We distributed the Android app to 105 PhoneLab participants for 45 days in an IRB-approved study and used camera monitoring of parking lots near Davis Hall to provide ground truth to evaluate our modeling algorithms. The map shown above displays some of the arrival and departure events we were able to detect in parking lots near Davis Hall where blue is located. Overall, we found that PocketParker was able to determine parking lot availability with high accuracy. Our work was published at UbiComp'14 and reported on by the MIT Technology Review.