Distributed parking spot detection with on-board sensors
Spinolo, Philip Luke
MetadataShow full item record
Drivers searching for parking are significant contributors to congestion in urban areas. It has been shown that informing these drivers about available parking can help alleviate some of this congestion and thus reduce overall travel time and emissions. However, informing drivers about available parking requires up-to-date knowledge of the occupancy of parking spaces in the area. For certain situations with well-controlled entries and exits, like parking garages, this is a simple process. For more distributed parking, as in open parking lots or curbside parking, the current approach is to deploy sensors at each individual parking space. A more dynamic occupancy detection system may be possible using vehicle-borne sensors to check for open spaces. As vehicle technology continues to advance, capable sensors may even be natively equipped on some vehicles and trim levels, with no need for aftermarket kits. However, when using sensors that make distance measurements to determine whether or not a space is open, a secondary system must be able to check that a detected opening is a parking space and not an intersection or a bus stop or other area that cannot be parked in. In this thesis, a method for detecting openings using a late-model vehicle’s ultrasound parallel park assist sensors and then verifying that the openings are valid parking using basic map data in OpenStreetMap is described. An overview of parking guidance systems as well as relevant sensors is also provided. The system is then tested in two stages, first for the ultrasound sensors by themselves and then for the combined detection and validation system in three different parking scenarios around Atlanta. Results show that the system is effective at identifying opening parking spaces both on the street and in parking lots, though parking lots with angled spots and GPS accuracy are both challenges for the system.