A Tale of Two Cities

Smart Parking Solutions

Parking spaces in densely populated cities can be elusive, and between 15% and 30% of the traffic in large cities across the world is thought to be caused by motorists ‘circling’ to locate a suitable parking spot. Meanwhile, the process of paying for roadside parking is far from universal, with different cities taking different approaches. For example, cities across the United States tend to be governed by roadside parking meters, while some Swiss cities have specific free ‘blue parking zones’, which come with short time limits and heavy penalties for overstaying.

Roadside parking operates very similarly in two cities, located in China and Taiwan, that have implemented applications using AAEON products. However, while these particular cities are renowned for their innovative approach to urban development, the system they share can appear anachronistic. Motorists park in allocated roadside spaces; a parking attendant then patrols a set area, stamping a paper ticket containing the car’s parking location ID and vehicle license number every 30 minutes. Motorists then take this ticket to any convenience store and pay for the time they spent parked there.

While convenient for motorists, the efficiency of this system presents questions when you consider that the Taiwanese city has approximately 100,000 roadside parking spaces, meaning there is a lot of ground for a limited number of parking enforcement officers to cover. While the Chinese city only has around 70,000 roadside parking spaces, according to the city’s transport commission, up to 500,000 cars occupy its roads each night, demonstrating a need for both more parking spaces, and a way to effectively administer them.

Both cities have increasingly looked towards IoT solutions to parking as part of wider smart city initiatives. As a result, two companies utilizing AAEON products worked with each city government to propose applications to improve their respective roadside parking systems but dealt with the challenges presented by each city differently. The proposals submitted to each city government varied in how they would operate but had the matching aim of making parking more efficient, reducing emissions caused by ‘circling’ for parking spots, and ensuring compliance with roadside parking rules.

Empowering Applications in Taiwan

For the project in Taiwan, the biggest hurdle was finding a balance between power and connectivity. Powering its smart parking meters via integration with Taiwan’s national power grid would not be possible, due to potential instability and energy surges that would jeopardize the function and accuracy of the meters. However, using a lower voltage power source would limit the range of available AI-capable solutions. As such, the company had to look to power sources outside of the city’s existing infrastructure, but that could still power the application.

The challenge of connectivity was less complex, given AAEON’s expertise in producing edge AI solutions for a variety of vertical markets. However, considering the lack of a strong, stable power source, the computer chosen would need to be power-efficient to avoid interruptions to the function of the application, which could result in inaccurate data being collected.


The mechanics of Taiwan’s smart parking meter was comprised of two street-level cameras, mounted to bollard fixtures adjacent to each roadside parking space. Powered by AAEON’s BOXER-8221AI, and fitted with the NVIDIA® Jetson Nano™, the meters could determine the duration that a car had parked in a space using real-time license plate recognition.

Featuring the NVIDIA Maxwell™ GPU architecture with 128 NVIDIA CUDA® cores, the BOXER-8221AI is capable of operating multiple neural networks or processing several high-resolution images at the same time, while maintaining ruggedness and energy-efficiency, providing the perfect balance for Taiwan’s smart parking meter application.

To process the parking data, each unit’s built-in camera would capture the vehicle’s license plate upon arrival and departure, along with the parking location, then send this data to the cloud. With this data, the company’s software would then calculate the fee due and send a text message to the motorist with the amount payable.

All Systems Go! - Infrastructure Solutions with Smart Collaboration

As previously stated, Taiwan’s smart parking meter required an independent power source, so the company tasked with creating the application looked towards a convenient, environmentally friendly solution, powering the smart parking meters using long-life, electronic scooter batteries.

A Taiwanese company specializing in the sale and rent of urban electric scooters powered by swappable and rechargeable modular batteries was happy to collaborate with the project. With 2,274 electronic scooter battery charging stations across Taiwan, which allow customers to swap out old scooter batteries for them to be recharged, the technology required to power the smart parking meters was already very prevalent across the city.

By manufacturing the smart parking meter mechanism to be compatible with swappable batteries, the company did not need to install any additional power lines or invest in additional infrastructural development. Compounding this benefit was that the BOXER-8221AI offered AI capability via the Jetson Nano, while its power consumption was minimal and so each battery was able to power a smart parking meter for 14 days before requiring replacement.

Connectivity - An LPWANderful Solution

The need for power-efficient hardware limited the company’s connectivity options somewhat, as modules with more sophisticated edge AI capacities are generally more power-consumptive. However, with the license plate data being identified by the AI inferencing capability of the Jetson Nano, and the application storing the camera’s live feed metadata and cost calculation on the meter at ground level, the only data that needed to be transferred to the cloud was the license plate number, duration of stay, and the parking fee.

As a result, the volume of data requiring transference to the cloud was reduced to only the images and times showing a vehicle occupying the parking space, meaning this could be sent to the cloud via CAT M1, a low-power wide area network (LPWAN) cellular technology, making the process power-efficient and aligned with the capabilities of both the Jetson Nano and the batteries powering the application.

China’s Challenges – The Innovation of an Historical City

Like in Taiwan, challenges to installing smart roadside parking in the Chinese city also focused on how the application could be powered, and how the application would process and transfer data. The primary point of difference between the two cities is the sheer number of vehicles involved.

With an estimated 5.4 million vehicles on the roads of this Chinese city, it would be less practical to implement the same model as in Taiwan, with two camera bollards per parking space. In addition to this, the Chinese city suffers from a potent combination of a large number of cars, and a rather unique topographical architecture.

While it is a high-tech metropolis, this particular Chinese city is also renowned for the cultural legacy of its geography, with the city’s landscape combining an ultramodern skyline with narrow alleys known as hutongs. It is consequently very difficult to install two car-level roadside parking meters per parking space without major infrastructure changes, which in many cases would be prohibited due to the historical value these areas hold.

China’s Solution – Six in One!

In a slightly broader approach, the solution presented for the Chinese city was to affix the application hardware to existing streetlight frames approximately 6 meters high, which would allow the integrated high-resolution camera to cover between four and six roadside parking spaces simultaneously. Therefore, the company bidding for the project chose the BOXER-8230AI embedded edge AI system, equipped with the NVIDIA® Jetson™ TX2 NX module, as it would provide the AI inferencing and processing power for such a task. With this simultaneous AI inferencing technology, the application could recognize and calculate parking fees for up to six cars across six spaces.

The proposed plan was that the application’s camera would record the duration of a vehicle’s stay in any given space based on license plate recognition, automatically calculating the cost using the company’s own software. The total amount payable would then be sent to the motorist via text message.

Getting on the Grid

The proposed plan to power the Chinese city’s smart roadside parking system was remarkably simple due to the city not having the same infrastructural issues as Taiwan with respect to its electricity grid. Rather than requiring an independent power source, the application could simply utilize the city’s grid power, without the need to build additional power lines. This would reduce the time and labor required for installation considerably, given the BOXER-8230AI’s wall mount kit.

Because of this, a greater source of power was available, which in turn enabled the city to use this increased power source on more sophisticated AI inferencing technology. Consequently, this would enable the application to monitor more parking spaces per camera, and reduce the physical number of units required to achieve the same result.

Connectivity – Parking on the Edge

With a boost in available power, and therefore inferencing capability, the smart roadside parking system was able to process a greater amount of data on the edge per unit, resulting in a highly efficient mode of operation. The unit’s high-resolution camera would photograph the license plate information of cars occupying up to six parking spaces, which could be analyzed and processed by the Jetson TX2 NX. A copy of the picture taken would be retained on the edge, within the smart roadside parking system unit. The application’s AI software would then generate a code representing this data and send it to the cloud.


The introduction of 500 smart parking meters in the Taiwanese city was very successful and was well received. The time spent looking for a parking space was said to have been reduced by as much as 43%. In addition to this, the ease and efficiency of the smart parking meter saw overall traffic volume decrease by approximately 8%, overall traffic speed increase by 3%, and saw parking availability increase in the area of deployment. By choosing AAEON’s BOXER-8221AI, the city government benefitted from not only the aforementioned benefits, but also an increase in parking revenue due to the reliability of the application, and the virtual impossibility of non-compliance, with a reduction in total overheads due to the BOXER-8221AI’s low power consumption. As a result, the project will be expanding, with an expected deployment of a further 1,000 spaces covered by smart parking meters by the end of the year.

While the company tasked with equipping the Chinese city with a smart parking alternative only required AAEON to provide proof of concept with the view of future implementation, it was demonstrated that the application powered by the BOXER-8230AI would be hugely beneficial and fully functional.

With the BOXER-8230AI’s compact size and powerful inferencing capabilities, along with the ease with which an application featuring it could be connected to existing infrastructure, such an application has incredible potential. Add to this the additional revenue stream that the city could gain from the automatic calculation and deduction of parking fees, and the reduction of overheads such as human parking enforcement officers, and it does seem that smart parking can contribute far more than just convenience.

For more information on AAEON’s BOXER-8230AI or BOXER-8221AI, please visit our product pages.