Whether you want to reduce congestion, increase parking revenue or reduce occupancy – or a mixture of all three – there is plenty of technology available. Andrew Bardin Williams considers the pros and cons.
Drawn in by the promise of Smart City initiatives, communities across North America are embracing smart parking solutions in an effort to change citizens’ transportation behaviours for the better. They are doing this by using policy and ITS solutions to help de-incentivise parking for most people while making it easier for those who have no other option. And, in some cases, it seems to be working – but in others, not so much.
According to a panel of parking experts interviewed for this article, recent smart parking deployments in Los Angeles, San Francisco and Washington, DC, have had varying levels of success - and those differences can be directly related to the availability of newer occupancy detection technology. As each deployment came up, new advancements in sensor and camera solutions improved accuracy and efficiency, allowing each successive deployment to consist of leaner networks that rely on powerful data analytics and algorithms. Essentially, new sensor technology enabled the smart in smart parking.
Role of occupancy detection
According to Mark Zannoni, a transportation analyst for IDC, smart parking solutions have four requirements. They must:
- identify or forecast open parking spaces and relay that information to drivers
- support multiple payment options - whether by meter, kiosk or smart phone
- support enforcement efforts—either by informing agents of expired meters or by embedding automatic enforcement
- feed valuable data to the city’s transportation agency to inform greater transportation policies and programmes such as traffic management and variable pricing initiatives
Occupancy detection can be done several ways. Sensors embedded in the roadway can use magnetometers or radar to identify vehicles. Fixed overhead cameras use vehicle detection software to identify open spaces. And mobile detectors can be mounted on trucks or vans. The detection data from each type can be plugged into an algorithm to fill coverage gaps, extrapolate data and predict the likelihood of other open spaces. Data is constantly updated so parking authorities can continually tweak and refine policies.
Cameras vs. in-ground sensors
Each detection technology has its own pros and cons. Overhead cameras can monitor multiple spots at a time - often an entire lane of street parking or a large area of a parking lot or structure. They can also help with enforcement if equipped with licence plate detection capabilities or vehicle type recognition, and they can identify other parking infringements such as double-parking and fire hydrant infractions. Most importantly, however, according to Julie Dixon, a parking consultant in Los Angeles, cameras can be used for other ITS purposes such as surveillance and traffic management – and their versatility can be a better fit in municipalities’ overall smart city initiatives. On the other hand, cameras are more expensive than in-ground sensors, require power and may be inhibited by weather, foliage, vibrations or vandalism. Video analysis technology is also more nascent than sensor technology, making it a risky investment for budget-strapped municipalities.
In-ground sensors have been around for decades and are deployed under pavements, where they are impervious to weather and line of sight issues. They have typically relied on magnets to detect metal in a parking space, but more accurate radar-equipped sensors are now available. Long battery life and simple, single-purpose engineering reduce maintenance costs - though if a sensor does need maintenance, it likely requires a crew to rip up pavement. As for those smart city initiatives…
Unfortunately, most in-ground sensors are unable to provide much data beyond parking space occupancy - making them primarily single-purpose deployments with limited ROI. So, what occupancy detection technology is better? The answer may be one, or both - or even neither. It’s all about reducing the city’s reliance on fixed infrastructure while still getting an acceptable insight into space availability.
Lessons learned in DC
While early deployments in Los Angeles and San Francisco were plagued by reliability, accuracy and scalability issues (see box, Inauspicious start in California), the most recent variable pricing programme in Washington has worked. The difference, according to Soumya Dey, assistant director of transportation operations and safety for District Department of Transportation (DDOT), has been time and good old engineering. In the years between the LA and Washington deployments, sensor technology had become a lot better and new camera solutions have entered the marketplace - helping to drop prices.
The first thing DDOT did was to analyse what went wrong with the smart parking deployments in Los Angeles and San Francisco and learn from the mistakes made. Immediately, it became apparent to Dey that the price of occupancy detection was a major inhibitor to the success of those deployments and that success in DC would have to be more cost-efficient. While Los Angeles and San Francisco deployed close to one sensor per parking space, DC would insist on a leaner one sensor to two spaces deployment. “We were okay with lowering our accuracy because we felt that we could make up for it with better analytics and predictive algorithms,” Dey said. “If we were right, we’d immediately save 50% on opex costs.”
DDOT also wanted to temper expectations by forgoing real-time variable pricing in favour of quarterly or monthly price changes. The department also made sure to set clear objectives. According to Dey, DDOT made it a goal to maintain one available space per city block at all times and prioritised congestion reduction in an area where 30-35% of traffic is due to drivers looking for a place to park. DDOT then conducted a six-month pilot programme across various neighborhoods to determine exactly how drivers would react to the district’s new variable pricing scheme. Six vans equipped with vehicle detection cameras were set up on various blocks for a week at a time, collecting valuable data that could help engineers come up with the variables they would need to figure out the probability of occupancy. By the time the variable pricing programme launched in 2016 with 500 sensors across 1,000 spaces, the city had a good idea of how drivers would react to different pricing and how that behaviour would affect congestion.
“You can’t manage what you don’t know,” Dey said. “We treat our curbsides as assets and maintain visibility into the state of that asset. We determined that we need to know who our users are, the turnover and how much revenue is tied to that particular space. Once we know that, we can formulate policies and push information to the public.” Every indication is that the DC programme has been a success.
The programme’s fifth price adjustment was made in October, and DDOT says it is meeting its goal of one available space per city block. In addition, a new congestion study is expected to come out in February (too late for publication in this article) but Dey hints that congestion has been reduced in the target area.
Making the right decisions
What’s interesting about the DC smart parking deployment is that DDOT was not afraid to use multiple sensor and camera technologies for different usages. Mobile cameras were ideal for DDOT’s pilot programme and provided a solid baseline from which to determine occupancy need and driver behaviour. Later, in-ground sensors provided the most bang for the buck in the permanent installation while cameras affixed to light poles fill any gaps in the sensor network. According to Dey, these decisions were based on DDOT’s foresight to learn from past mistakes and clearly define its goals so subsequent technology decisions could address specific needs.
The experts interviewed for this article all advised municipalities that are considering smart parking solutions to slow down and really think about the goals that the programme has and how accurate the data collected needs to be. Do you want to reduce congestion? Increase parking revenue? Reduce occupancy? De-incentivise single-occupancy vehicles? Eliminate meter enforcement personnel? “You really need to define your objectives,” Dixon said. “Smart parking is not just something you can plug in. It can be expensive to deploy and to maintain. You need to determine exactly what you want to get out of the programme.” As occupancy detection technology continues to evolve, municipalities are going to have tough purchasing decisions. They would be best served by determining the specific needs of the deployment and going from there.
Inauspicious start in California
Los Angeles rolled out a smart parking programme called ExpressPark in 2011 across 7,000 downtown street parking spaces. The deployment relied on magnetometers to detect whether spaces were occupied and sent that information to a central management console where it could be monitored by transportation officials. The goal was to direct drivers to available spaces and vary pricing according to demand. According to Arti Gupta, a parking consultant for LA and other municipalities, the rollout didn’t go exactly according to plan.
Reliability and accuracy
Using a one sensor for one parking space ratio, the project become unscalable past the 400-space pilot area, and sensors began interfering with one another. Data accuracy, promised to be 95%, hovered much lower than that, preventing the public from trusting the information they were getting from the system. Around the same time, a smart parking programme in San Francisco was having similar issues. According to Dixon, who worked on the project, magnetometer sensors that were deployed in each space were being affected by the city’s extensive bus and light rail systems, causing major reliability and accuracy issues. Officials in both Californian cities worked with vendors to fix the issues, but the programmes have yet to recover fully from their rocky starts.
