3D vision aided GNSS real-time kinematic positioning for autonomous systems in urban canyons

A method for estimating a position of a vehicle, the vehicle being configured to receive navigation signals of a global navigation satellite system (GNSS), the method comprising:

• obtaining a first set of GNSS data derived from a first plurality of navigation signals received by the vehicle for positional measurement, wherein the first plurality of navigation signals excludes any navigation signal identified to travel from the GNSS to the vehicle via a non-line-of-sight (NLOS) path, allowing the first set of GNSS data to be obtained from healthy but high-elevation satellite measurements;
• obtaining a second set of GNSS data derived from a second plurality of navigation signals received by a reference station for providing real-time kinematic (RTK) positioning in vehicle-position estimation;
• generating a plurality of GNSS factors indicative of a contribution of GNSS measurement to vehicle-position estimation according to the first and second sets of GNSS data;
• identifying a set of low-lying visual landmarks from a first set of images that capture low-lying environment structures visible to the vehicle for visually measuring a near-ground environment;
• generating a visual factor indicative of a contribution of visual measurement to vehicle-position estimation according to the set of low-lying visual landmarks;
• optimizing a plurality of system states used in a factor graph for integrating measurement data obtained in plural measurements to yield a float solution of the vehicle position, the factor graph being formed by a plurality of factors, an individual factor being indicative of a contribution of a measurement to vehicle-position estimation, wherein the plurality of factors includes the visual factor and the plurality of GNSS factors, thereby utilizing complementariness between the set of low-lying visual landmarks and the first plurality of navigation signals for improving geometry constraints in vehicle-position estimation to enhance estimation accuracy;
• and
• resolving integer ambiguity of the float solution to yield an estimated position of the vehicle.