Vision Aided Inertial Navigation

Today GPS aided inertial navigation is widely used and well studied in any aspects. The good short term properties of inertial data are complemented by the long term stability of the GPS signal. A common approach is to use a Kalman filter for fusing GPS and inertial data to constrain the inertial sensor drift. Although this is working well in many applications there is a need to find a similar solution for navigation tasks in difficult environments with erroneous or no GPS data. Therefore a vision aided inertial navigation system is presented which is capable of providing local navigation for indoor applications without GPS or could be used to bridge GPS dropouts in urban or forested areas. A method is described to reconstruct the ego motion of a stereo camera system aided by inertial data which in turn is used to constrain the inertial sensor drift. The optical information is derived from natural landmarks which are extracted and tracked over consequent stereo image pairs. Using inertial data for the feature tracking effectively reduces the computational effort and the uncertainties from mismatching by allowing smaller search areas. This is an important precondition for a robust tracking algorithm running in real time. However, before fusing the data with a Kalman filter many aspects including synchronization and determination of alignments of the sensor systems as well as the calibration of the stereo camera system have to be considered. The results of an integration of optical and inertial navigation are shown on an indoor navigation task.