Research On The Dynamic Update Algorithm Of GIS Based On Environmental Perception Of Autonomous Car

Driverless car is a hotspot in mobile robot study. Autonomous navigation is importantfor driverless cars. Accurate autonomous navigation is based on the accurate environmentalmap. Geographic Information System (GIS) can provide an accurate map, and plan pathsfor an driverless car. Thus it is significant to establish GIS for driverless cars. With the helpof GIS, driverless car can drive safely and quickly.For the research on the dynamic update algorithm of GIS based on perceptioninformation, this dissertation mainly focuses on three-dimensional localization ofperception object, the vehicle positioning algorithm and dynamic update of GIS based onleast square method (LSM). The mainly studying results of this dissertation are as follows:Firstly, the perception information is acquired from3D laser radar and cameras, andthree-dimensional localization is applied to get perception object position relative to theautonomous car. Perception particle model is constructed, which represents perceptionobject. Three-dimensional localization is based on laser perception and visual perception.Secondly, a positioning algorithm is proposed in this dissertation, which is based onmap-matching and environmental perception for autonomous car. Macroscopically, thealgorithm makes use of computational geometry to match the position of autonomous car tothe corresponding road, based on GPS point and map information of the road network.Microscopically, it makes use of the environmental perception to make precise positioning.Finally, based on three-dimensional localization and positioning algorithm, thisdissertation proposed a dynamic update algorithm of GIS using cursive least squaresmethod (RLS), which adds perception particle and perception area into GIS, to realize thedynamic update of GIS based on perception information. RLS estimates the location ofperception particle to reduce the random error.Through practical car test, the success rate of positioning algorithm is99%,positioning time for a single point is less than2milliseconds, and the positioning accuracyis about1meter. The perception particle and perception area are added into GIS, realizingthe dynamic update for GIS based on perception information.