Deteção de lugar de estacionamento com dados de sensor

The daily search for a parking space eats up around 560 million hours a year in Germany.

Comsysto Reply has developed a service for parking space detection as part of the driver assistance functions for an innovative global automobile manufacturer. The customer required a system for parking spaces detection based on data collected via ultrasound sensors on the vehicle. The data are filtered from several terabytes of log data from the test fleet and then aggregated. Statistical models are created based on the pre-analysed data.

A further objective was to set up a highly scalable data analysis platform for batch and streaming data analysis. The efficient storage and retrieval of georeferenced data was another of the project's challenges.

Comsysto Reply opted to use an exploratory approach with rough targets to develop the parking space detection, which is part of the driver assistance as standard. The solution visualises the data in real time, and checks whether the data gathered are suitable for the models. The services are provided via REST web services (microservices). The GeoJSON standard is used so that the data gathered can also be easily used by other systems. Comsysto Reply used automated configuration and administration of the IT infrastructure - quality assured for AWS or internal servers.

Advanced Analytics and Data Science

By using Advanced Analytics and Data Science, users benefit from visual and statistical data exploration with data science notebooks. Missing figures such as GPS data are added using data curation. The solution uses feature extraction to provide the prediction model for the sensor data and make it operational. Comsysto Reply uses statistical models for a variety of services in order to continuously improve the service for those looking for parking spaces. Another key part of the project was the discretisation of the sensor and geodata for the aggregation.

The service has been in use since March 2017 for selected vehicle models in ten German inner cities (Berlin, Bremen, Düsseldorf, Frankfurt/Main, Hanover, Hamburg, Cologne, Munich, Nuremberg, Stuttgart) and helped drivers find a parking space.

The service displays the parking space situation at the destination in real time and shows the customer the probability of finding a parking space.