February 2, 2018

August 8, 2017

Please reload

Recent Posts

Parking & Drones: A Match Made in Heaven?

August 19, 2019

1/4
Please reload

Featured Posts

Sensors vs Analytics: How to Find Open Parking Spaces

April 28, 2017

 

Summary: using existing technology, there are two methods that can be used to help motorists locate available parking without causing congestion and pollution by circling the block.

 

Key points:

  1. Predictive analytics use past data and provide parking locations that are likely available

  2. Sensors provide real time data on open spots, updating almost instantly

  3. Scroll to the end for a chart comparing the two options

 

Introduction

 

When you think of industries that are impacted by technological advancements, parking generally doesn’t come to mind. Fair – from the invention of the parking meter in 1935 through the following 60+ years, basically nothing changed[i]. However, parking started to evolve with the invention of electronic, multi-space meters in the 2000’s. And with the popularization and wide spread nature of smartphones, an explosion of parking apps that let you pay for parking with your smartphone have hit app stores around the world.

 

One thing we haven’t managed to demystify yet is how to know exactly where there’s a spot available when you need it. Given currently available technology, this conundrum has two possible solutions: predictive analytics and sensors. In this article, you’ll learn the difference between the two, and how parking is benefiting from both options. So without further ado, let’s see how we can find an open parking space when and where we need it!

 

 

Predictive analytics

 

Predictive analytics is the concept of collecting huge amounts of historical and transactional data, sorting it, and using patterns found in that data to identify risks and opportunities for the future[ii]. It captures relationships among many factors to determine the probability (high, medium, low) of something occurring in a specific set of circumstances (like time, day of the week, etc.).

 

Applying this to parking, you could take a map of an area that parking is permitted, and colour code streets based on the historical likelihood of a spot being either empty or taken at that point in time. Here’s a great example of where a picture is worth 1000 words (or about 2500 if you’re as terrible at describing something as I am)…

 

 

 

Basically you’re seeing that under a specific set of circumstances (Mondays at 1:00pm), there are many parking spaces available on the corner of 4 Ave and 5 St SW in downtown Calgary, so the probability of something occurring (finding a parking spot in this area) is high.

 

Now the example above requires manual prompting of the system to gather the data, and then a human has to look at the results and take action based on them. This isn’t very practical when you start using predictive analytics on a larger scale (say for controlling traffic during rush hour). To benefit from predictive analytics on a larger, automated scale (like having traffic lanes act as one way out of the downtown core from 4-6pm, and notify motorists accordingly), we use prescriptive analytics. It’s the term used to automate these complex decisions and generate actions based on the analysis. It’s not involved in parking yet, but I’m sure you can imagine how autonomous vehicles will be able to take full advantage and find the spot with the lowest price and highest likelihood of attracting another passenger in the shortest amount of time.

 

Let’s backtrack for just a minute before we move on. There are a ton of applications for this concept, including fraud detection (credit applications, transactions, identity theft, insurance claims), health care (identify at-risk patients), marketing (cross selling opportunities, types of ads to show, identifying prospects), and many more[iii]. An ensemble model can also go further and use the ‘wisdom of crowds’ technique to pool the predictions of many models and generate an even more accurate answer or best practices.

 

The reason this is important to us in the parking world is that there are huge potential cost savings. Imagine knowing exactly who was going to be influenced by the ‘park here’ signs you put up, and not bothering to show them to people who weren’t. Or who would become a loyal customer with a bit of encouragement, and who never would? You could save a lot of money if you knew exactly who to target as parking clients[iv]. Ah, so many possibilities!

 

If I lost you somewhere along the way, here is a final offering to reinforce the concept of how cool predictive analytics are.

 

Sensors

 

The other way technology can help find you a parking spot is through sensors. There are many types of sensors, including speed, temperature, pyroelectric, optical, chemical, environmental, etc[v]. They all work in a similar way, with each individual sensor (or slave) collecting data and, via a connection with a hub (master or anchor), the transmitting the information to the storage and processing location.

 

The ones we’re interested in here are those that have an application within parking. Some are great for moving vehicles (proximity, photoelectric), while others are more useful stationary vehicles (pressure). Here’s the low-down:

 

Proximity: a familiar example here is a backup camera, which senses how close your vehicle is to another object and alerts you based on distance.

Photoelectric: we use this technology a ton in parking lots and garages; it’s used for detecting entries and exits of a controlled area. It can also tell you the number of inbound and outbound vehicles on any given section of street, allowing an educated estimate of available parking[vi]. For example, if there are 10 parking spots on this section, 10 cars enter, and 0 cars leave, well the odds of the 11th car find a parking spot are likely zero.

Ultrasonic: another important one in parking, we rely on this type of sensor to make sure gates don’t close on top of cars[vii].

Pressure: like changing a light from green to red when a car is parked on top of a sensor, which is exactly what we do in parking guidance systems.

 

Regardless of what type of sensor is installed, over time they all use machine learning to improve their recognition, identification and analysis skills. This makes for better quality, more accurate information, and therefore better results[viii].

 

So how can any of these sensors help us find a parking spot – in real time?

 

Well, going back to the pressure sensor, it’s pretty clear how identifying a spot by colored light bulb can help us find a parking spot in any given enclosed lot – once you’re already in the lot. Which is great when you’re at the mall on the Saturday before Christmas trying to get your shopping done without wasting 2 hours trying to find a spot. But what about when you’re thinking about driving downtown for an important meeting and you don’t know where you’ll be able to leave your vehicle?

 

Here’s where pressure sensors collide with smart city technology (which you can learn all about in a previous post here). By installing sensors under individual parking spots, either on-street or off-street, a live feed of the status of that parking spot (occupied or vacant) can be shared and accessed by the public. This can be done via an app or a public website and produce the same results. And if you’re particularly considerate, as an operator you can even offer the ability to reserve that spot and pay for it in advance.

 

Once you start linking all these individual pieces of information together to build a bigger picture, you could see what spot is occupied, and also detect if an associated payment has been made if in fact it is in use. As a parking operator, this information is of great value, as you can schedule officers on the most efficient routes, or deploy them on an as-needed basis. You can price out the cost of sending an officer and weigh it against the value of infractions and fines they are going to hand out. Or you could map out a route for the officer to take that avoids areas where no cars are parked, thus using time as efficiently as possible[ix].

 

 

Which Method is Better?

 

Well each one has advantages, so it really depends on the goals of installing the system and the resources available. To help with a quick comparison, take a peek at this chart:

 

 

Still want more info?

 

There are several great scholarly articles on Sensors journal – both for parking applications and other uses. For a more anecdotal take on which is better, read Richard Simpson’s blog post. You can also be swayed toward the sensor camp by the