A Novel Subspace Method for Robot Localization Using Landmark Bearings

We propose a subspace approach to solve the robot localization problem from landmark bearings. The proposed method can utilize all bearings and the inter-landmark bearings to jointly estimate robot location and its orientation. Instead of deriving pseudo-linear equations, the proposed method improves the performance of robot localization by making use of the dimension and eigen-structure of the multidimensional similarity matrix. Simulation results demonstrate the effectiveness of the proposed algorithm.