A smart parking system can prevent frustration and emissions MIT News

It happens every day – a cross-town driver checks a navigation app to see how long the trip will take, but finds no available parking when he arrives at his destination. By the time they park and walk to their destination, they have traveled much further than they expected.

Popular navigation systems send drivers to a location without considering the additional time it may take to find a parking space. This causes more than a headache for drivers. It can worsen congestion and increase air pollution by causing drivers to drive around looking for a parking space. This stigma can discourage people from taking mass transit because they don’t realize that it can be faster than driving and parking.

MIT researchers have tackled this problem by developing a system that can be used to identify parking spaces that offer the best balance of proximity to a desired location and the likelihood of parking availability. Their flexible approach directs users to the ideal parking location rather than their destination.

In simulation tests with real-world traffic data from Seattle, this technique achieved time savings of up to 66 percent in very congested settings. For a motorist, this will reduce travel time by about 35 minutes, compared to waiting for a space to open up in the nearest parking lot.

Although they have not yet designed a system ready for the real world, their demonstrations show the effectiveness of this method and show how it can be used.

“This frustration is real and felt by many people, and the big problem here is that systematically underestimating these driving times prevents people from making informed decisions. It makes it very difficult for people to make shifts on public transit, bicycles, or other modes of transportation,” said MIT graduate student Cameron Hickert, lead author of a paper describing the work.

Hickert was joined on the paper by Sirui Li PhD ’25; Zhengbing He, research scientist at the Laboratory of Information and Decision Systems (LIDS); and senior author Cathy Wu, Class of 1954 Associate Professor of Career Development in Civil and Environmental Engineering (CEE) and the Institute for Data, Systems, and Society (IDSS) at MIT, and a member of LIDS. The study appears today on Developments in Intelligent Transportation Systems.

Possible parking

To solve the parking problem, researchers developed a probabilistic recognition method that considers all possible public parking spaces near the destination, the driving distance from the destination, the walking distance from each location to the destination, and the probability of parking success.

The method, based on dynamic programming, works backwards from the best results to calculate the best route for the user.

Their method also considers the situation when a user arrives at an ideal parking spot but cannot find a space. It takes into account the distance to other parking spaces and the probability of successful parking in each space.

“If there are a few lots nearby that have a slightly lower chance of success, but are very close together, it might be a smarter play to drive there than to go to a high-probability spot and hope to find an opening. Our framework can account for that,” Hickert said.

Ultimately, their system can pinpoint the ideal location with the least expected time required to drive, park, and walk to your destination.

But no motorist expects to be alone trying to park in a busy city centre. Thus, this method also includes the actions of other drivers, which affect the user’s chances of successful parking.

For example, another driver may arrive at the user’s preferred location first and take the last parking spot. Or another driver can try to park in another place but park again in the convenient place of the user if he does not succeed. In addition, another driver may park in a different location and cause a flow of results that lowers the user’s chances of success.

“With our framework, we show how you can balance all those factors in a clean and systematic way,” Hickert said.

Crowdsourced parking data

Parking availability data may come from several sources. For example, some parking lots have magnetic sensors or gates that track the number of cars entering and leaving.

But such sensors are not widely used, so to make their system deployable in the real world, researchers have learned the efficiency of using full-source data instead.

For example, users can indicate available parking space using the app. Data can also be collected by tracking the number of cars circling to find a parking space, or how many enter and exit after failing.

One day, autonomous vehicles may even report open parking spaces in which they drive.

“Right now, a lot of information doesn’t go anywhere. But if we can capture it, even if someone just taps ‘no parking’ into the app, that can be a valuable source of information that allows people to make informed decisions,” Hickert said.

The researchers tested their system using real-world traffic data from the Seattle area, simulating different times of the day in a congested city and suburban area. In congested conditions, their method reduces travel time by about 60 percent compared to sitting and waiting for a space to open up, and by about 20 percent compared to the strategy of continuing to drive to the next parking lot.

They also found that people’s perception of parking availability would have an error rate of only about 7 percent, compared to actual parking availability. This shows that it can be an effective way to collect data on parking opportunities.

In the future, researchers want to conduct larger studies using real-time traffic information across the city. They also want to explore more ways to collect data on parking availability, such as using satellite imagery, and measuring potential pollution reductions.

“Transportation systems are so big and complex that it’s really hard to change them. What we’re looking at, and what we’ve found this way, are small changes that can have a big impact on helping people make better decisions, reducing congestion, and reducing emissions,” said Wu.

This research was supported, in part, by Cintra, the MIT Energy Initiative, and the National Science Foundation.