Parking has become a challenge due to appearance of more and more cars. The city-wide parking guidance system (CPGS) is based on real-time parking data throughout the city. Data collection costs extremely high, thus an effective parking guidance system is needed.

A research team led by Dr. PENG Lei from the Shenzhen Institutes of Advanced Technology (SIAT) of the Chinese Academy of Sciences has proposed a new lightweight technical solution, addressing the current situation for lacking parking data.  

Dr. PENG has built a model based on the public parking information authorized by the government of Shenzhen city. The model parameters include capacity, service scope, location, etc.  

Compared with other counterparts, this model didn’t need the complete real-time data. As for the secondary data used in this model, the research team has employed an automatic data generation method named Generative Adversarial Networks (GAN) to fit the missing data according to the spatiotemporal similarity among parking, which reduced the cost of data collection from city-wide perspective.  

The research team furthermore developed a driver-friendly vehicular APP with a built-in model for recommending a proper parking place, considering factors such as destination, driving context, and estimated parking possibility.  

The newly developed APP could provide the driver a dynamic, personalized and supreme option via voice and Graphical User Interface (GUI).  

"We have spent six months on collecting data, training model and carrying out the parking guidance experiments. Compared with traditional methods, the time cost of the new system is reduced up to 70%. And the guidance effect is obvious, especially in rush-hour and hot-spot areas. The time spent on cruising for parking is reduced by 20% on average," said Dr. PENG. "We are looking forward to cooperating with industrial partners to push forward the technical solution to commercial product." 

The study was published in IEEE Intelligent Transportation Systems Magazine.

"In my opinion, the technical solution presented in this paper for improving parking guidance service looks promising. Also, I highly appreciate that the work is implemented and tested in a real scenario," said Porf. Vlacic Ljubo, chief editor of the journal.