Probabilistic Mapping for Unmanned Rotorcraft Using Point-Mass Targets and Quadtree Structures
Derek S. Caveney, Yeonsik Kang, J. Karl Hedrick · 2005
In this paper, the authors present a technique for constructing probabilistic occupancy maps for unmanned rotorcraft. The mapping technique is Bayesian and assumes that a ranging sensor positioned on the rotorcraft is providing noisy target measurements in the presence of clutter. By running a multiple-model Kalman filter-based algorithm, all measurements are used to provide point-mass target state estimates and associated state covariances. Furthermore, the multiple-model algorithm provides probabilities that indicate whether each target is a true target or a false alarm. These three attributes of each target (the state, covariance, and true target probability) are used to build and update the occupancy map. The map is built upon a quadtree structure that allows for higher map resolution in occupied areas of the operating environment and quick access of obstacle locations. This aspect of a quadtree-structured map is particularly useful when transferring obstacle information to obstacle avoidance and route planning routines. The construction of occupancy maps resulting from this quadtree-based, probabilistic technique is demonstrated through simulations of a low-flying rotorcraft travelling through an urban landscape.