It happens every single day: A driver driving across town checks a navigation app to see how long the journey will take, but when he gets to his destination he finds that there aren’t any parking spaces available. When they finally park and walk to their destination, it is way later than expected.
Most common navigation systems direct drivers to a location without taking into consideration the extra time it might take to seek out parking. This causes greater than only a headache for drivers. It can worsen congestion and increase emissions as drivers must drive around on the lookout for a automobile parking space. This underestimation could also discourage people from using public transportation because they don't realize that it might be faster than driving and parking.
MIT researchers addressed this problem by developing a system to discover parking spaces that supply the most effective balance between proximity to the specified location and likelihood of parking availability. Their adaptive approach guides users to the best parking spot, not their destination.
In simulated tests using real traffic data from Seattle, this technology achieved time savings of as much as 66 percent within the busiest areas. For a driver, this may reduce travel time by about 35 minutes in comparison with waiting for a free space at the closest parking zone.
Although they’ve not yet developed a practical system, their demonstrations show the feasibility of this approach and the way it could possibly be implemented.
“This frustration is real and felt by many individuals, and the larger problem is that systematically underestimating these travel times prevents people from making informed decisions. This makes it much harder for people to change to public transit, bicycles, or alternative types of transportation,” says MIT graduate student Cameron Hickert, lead writer of a paper describing the work.
Hickert is joined within the work by Sirui Li PhD '25; Zhengbing He, research scientist on the Laboratory of Information and Decision Systems (LIDS); and senior writer Cathy Wu, 1954 Career Development Associate Professor of Civil and Environmental Engineering (CEE) and Institute for Data, Systems, and Society (IDSS) at MIT and a member of LIDS. The research appears today in .
Likely automobile parking space
To solve the parking problem, researchers developed a probabilistic approach that takes under consideration all possible public parking spaces near a destination, the gap that should be traveled there from a start line, the gap that should be traveled from each automobile parking space to the destination, and the probability of parking success.
The dynamic programming-based approach works backwards from good results to calculate the most effective route for the user.
Their method also takes under consideration the case where a user arrives at the best parking spot but cannot discover a space. The distance to other parking spaces and the probability of success of parking in each of those parking spaces are taken under consideration.
“If there are several properties nearby which have a rather lower likelihood of success but are very close to one another, it is likely to be wiser to drive there slightly than go to the property with the upper probability and hope to seek out an open spot. Our framework can take that under consideration,” says Hickert.
In the tip, their system can discover the optimal property that requires the least expected time to drive, park and walk to the destination.
But no driver expects to be the just one attempting to park in a busy downtown area. This method also takes under consideration the actions of other drivers, which affect the probability of the parking process being successful.
For example, it might be that one other driver arrives first on the user's desired automobile parking space and takes the last automobile parking space. Or one other driver could try and park in a special automobile parking space, but then park within the user's ideal automobile parking space if unsuccessful. In addition, one other driver may park in a special automobile parking space, which can lead to negative effects that reduce the user's possibilities of success.
“With our framework we show how you possibly can model all of those scenarios in a really clean and principled way,” says Hickert.
Crowdsourced parking data
Parking availability data could come from multiple sources. For example, some parking lots have magnetic detectors or gates that track the variety of cars entering and exiting.
However, since such sensors will not be widely available, the researchers examined the effectiveness of using crowdsourced data to make their system more suitable for real-world use.
For example, users could display available parking spaces via an app. Data may be collected by tracking the variety of vehicles circling on the lookout for a automobile parking space, or what number of enter and exit a automobile parking space after being unsuccessful.
One day, autonomous vehicles could even report on free parking spaces as they drive by.
“Right now, loads of this information isn't going anywhere. But if we could capture it, even by someone simply tapping 'no parking' on an app, that could possibly be a vital source of knowledge that permits people to make more informed decisions,” Hickert adds.
The researchers evaluated their system using real traffic data from the Seattle area, simulating different times of day in a crowded urban area and a suburban area. In busy environments, their approach reduced overall travel time by about 60 percent in comparison with sitting and waiting for a parking zone to open, and by about 20 percent in comparison with the strategy of always driving to the following parking zone.
They also found that crowdsourced observations of parking availability had an error rate of only about 7 percent in comparison with actual parking availability. This suggests that this could possibly be an efficient strategy to collect parking probability data.
In the longer term, the researchers need to conduct larger studies with real-time route information across a whole city. They also need to explore other ways to gather data on parking availability, corresponding to using satellite imagery, and estimate potential emissions reductions.
“Transportation systems are so large and sophisticated that they’re really difficult to alter. What we’re on the lookout for, and what we now have found with this approach, are small changes that may have a huge impact to assist people make higher decisions, reduce congestion and reduce emissions,” says Wu.
This research was supported partially by Cintra, the MIT Energy Initiative, and the National Science Foundation.
