The ParkNet Project: An Intelligent Parking System


  • Vladimir Coric (vladimir.coric@…)
  • Patrick Yuen (pyuen2008@…)


  • Marco Gruteser
  • Ivan Seskar


Despite recent and significant investments to enhance and improve worldwide road infrastructure, the capacity of road networks has not kept pace with the demands for roadway efficiency of the growing driving populace. As a result of the dwindling supply of space, idle traffic congestion ,resulting from desperate drivers, has become an endemic to many bustling city and town officials who wish to facilitate their already chaotic highways and streets.

Especially in the densely populated areas, the effort to find the dwindling number parking spaces has worsened congestion and its debilitating effects to the economy,health, and busy schedules of members of the community. While searching for scarce parking spaces, frustrated drivers are slowing down traffic,increasing travel times of travelers, and wasting the time of many desperate working citizens. However, by utilizing the ParkNet System, this problem can be remedied by providing needy drivers with a reliable real-time generated map with information of street conditions and parking spot availability.

Project Description:

ParkNet is a mobile system originally developed by Suhas Mathur, Tong Jin, Nikhil Kasturirangan, Janani Chandrashekharan, Wenzhi Xue, Marco Gruteser,and Wade Trappe (Click here for a link to the award winning paper)comprised of multiple scouting vehicles (such as ubiquitous taxis) that survey and collect parking space occupancy information. In the ParkNet system, "sensor" vehicles are equipped with ultrasonic rangefinder sensors. While the vehicles travel around the area, the sensors will constantly determine the occupancy status of the various parking spots within the region. As the information is collected, the data is simultaneously relayed with data from an onboard GPS device which reports location along with the sensor's collected availability information to a central server. This summer, the project hopes to create an easily decipherable map from the incoming data which users can utilize to find elusive parking spaces.

Data set:

The Parknet data set is collected from previous tests done in downtown Brooklyn using 6 different ParkNet sensor vehicles. The data consists ~1.5 million data points, each composed of a set of the testing vehicles' GPS coordinates and varying range values collected from the ultrasonic sensor. The ultrasonic sensor provides readings from 12 to 255 inches every 50 ms. In addition to coordinates and sensor readings, camera pictures are available from every data point. These camera pictures are not part of the system as they are used in experiments for the ground truth.

MATLAB Application

Throughout the project, this MATLAB application was an incredible asset for processing the near innumerable amounts of vital data. Generated in MATLAB by Temple Graduate Student Vladimir Coric, this application is a variation upon the GUI originally used by the first ParkNet Project. Originally intended for use in identifying and comparing the results from ParkNet ultrasound sensors and the ground truth generating cameras on top of the devices, this application required a user to manually input whether or not a car was in sensor reading (as identified as a line). The input was then recorded and compared to the collected data (small line graph on the side). However, the new application not only added variety to the original options(from: "no car and a car" to "no car, car, intersection, fire hydrant, wrong lane, etc...), it greatly reduced the time taken in analyzing the collected ~1.5 million data points.

Authors: Vladimir Coric (vladimir.coric@…), Patrick Yuen Advisor: Marco Gruteser, Ivan Seskar

Last modified 7 years ago Last modified on Aug 2, 2012, 4:38:19 PM

Attachments (12)

Download all attachments as: .zip