Never forget where you left your car again, because your phone knows! Also your car is probably worth thousands of dollars, so you should be keeping track of it anyway!

by worstideas

The issue:

When parking on the street or in an enormous shopping center parking lot, it can be easy to lose track of exactly where one’s car is parked.

Since cell phones constantly record a person’s GPS location as a standard feature (if you are not familiar with this, look up “iPhone Track location”—the images are quite striking), we can use this same data to reconstruct the car’s location when it was parked.


Fig 1: Your phone typically does not make this data easily available to you, but it is constantly recording (and saving) your location. This is a low-resolution zoom-out of tracking signals of a phone taken to Las Vegas. Each dot on the map actually represents dozens or hundreds of specific location “pings,” which are just not visible at this zoom level. The black point cluster is Las Vegas itself.


It would be useful if your phone could always tell you where your car was parked—without requiring any user interaction or planning ahead of time.

Luckily, this is possible!

The car location will can be inferred using two sources of data:

  1. By using the accelerometer of the phone (as a pedometer):
    1. When the user is driving, the pedometer should register no (or very, very few) steps.
    2. After parking, the pedometer should suddenly see activity.
  2. By examining the speed of travel between GPS coordinates.
    1. Data points that have an associated speed above 20 miles per hour are practically guaranteed to be in a car (or other form of motorized transportation).
    2. Car data points will still have interruptions (e.g. stop lights) and low-speed sections (e.g., traffic jam) that need to be accounted for.
    3. At some point, the driver will get out of the car and walk to their destination. This can be easily detected by the slower movement and non-zero pedometer data.

See figures 2 and 3 below for an example of integrating these two data types (top of figure = pedometer activity bar graph, bottom of figure = map and GPS “pings”). Try to figure out the parking spot on the diagram below.


Fig 2: Here is some fake sample data. The blue bars along the top (“Number of Steps Detected”) show pedometer / accelerometer activity from 9:02 AM to 9:10 AM (the more a person walks, the higher the bar). The yellow-to-orange-to-red rectangles at the bottom indicate the GPS locations at these specific times. Try to figure out where the user parked the car based on this data. See Fig. 3 for the algorithm’s guess.


Fig 3: Here is the algorithm’s guess for the parking spot—see if you agree with this guess! This is an annotated version of the data in Figure 2.


This feature should definitely be built into your phone!

PROS: Automatically lets you know where you (probably) parked your car, and doesn’t need any data that a modern cell phone isn’t already collecting.

CONS: Might not work very well in underground parking garages. Try to remember where you parked in those situations!