ParCaster searches for free parking spaces with AI

Students at the Eastern Switzerland University of Applied Sciences (OST) have developed the ParCaster. According to a press release, this is an AI tool that provides a forecast of parking space availability at 14 different locations in the city of St.Gallen.

The tool is based on a model developed by Rita Haffter and Jan Frei, graduates of the Machine Learning for Software Engineers program at the OST. ParCaster is now available to the public as a web version, making it easier for drivers to locate a free parking space in the city.

According to the press release, the underlying model uses data on weather, public holidays, events and school holidays to make predictions about parking space availability at 14 different locations in St.Gallen, including the train station, Kreuzbleiche, Neumarkt, Spelterini and Spisertor.

Haffter and Frei used publicly available historical parking space data to train a long short-term memory (LSTM). This is a “type of neural network that learns to recognize patterns from a series of information”, explains Haffter. “Through training, the LSTM is able over time to differentiate between important and insignificant information.”

Following a few setbacks – initially the model always showed the same free parking space – Haffter and Frei have been able to make the ParCaster available to the wider public. Users can enter the date and expected parking time in the tool and see in advance where they are likely to find free parking spaces. ce/gba