Study on the Uncertainty Propagation of Feature-Based Visual Navigation for Critical Flight Phases of Urban Air Mobility

As the demand for Autonomous Urban Air Mobility (UAM) has grown significantly, airborne navigation systems are facing more stringent requirements during critical flight phases such as vertical take-off and landing. However, it is difficult to simultaneously achieve high accuracy, integrity, and availability only using Global Navigation Satellite Systems (GNSS) in urban areas due to the presence of multipath and signal blocking effects caused by urban infrastructure. As a complementary solution, vehicle’s poses can be accurately estimated using feature-based visual navigation with a fiducial marker without the necessity of installing complex infrastructures. However, it is challenging to quantify the integrity performance of the feature-based positioning algorithm for aviation applications due to the inherent complexity of the system. One of the main challenges is to properly propagate the uncertainty from the measurement noise to the estimated positions. In our previous work, a six degrees of freedom Dilution Of Precision (DOP) for visual positioning was proposed to model the geometric impact of the estimation uncertainty, which uses the state-of-the-art Euclidean coordinates parameterization for the camera positions. However, it has been observed from flight campaigns that the Euclidean parameterization can only conservatively bound the uncertainty of the position estimates at very low altitudes. This paper presents a novel uncertainty propagation method based on parameterizing camera poses using the Lie group. The efficacy of this method is demonstrated through experimental results obtained in scenarios simulating the vertical operation of UAM. The results indicate that this method accurately reflects estimation uncertainty and outperforms existing state-of-the-art methods.