Localization and navigation are fundamental techniques for Biped Pole-climbing Robots (BiPCRs) to automatically perform tasks in spatial trusses. Since spacial trusses are representative self-similar and featureless environments and BiPCRs implement climbing by turning themselves upside down, existing localization methods often fail in such cases. This paper proposes a stepwise localization method for BiPCRs, which first coarsely estimates the pose of the base gripper by detecting a known trajectory using ultra-wideband ranging measurements, and then refine the coarse estimation by solving a two-dimensional geometric problem related to the robot linkages and trusses. Experiments show that the proposed method is able to locate a BiPCR in spacial trusses with a position error less than 100mm and an orientation error less than 5°. The proposed method provides the BiPCRs with a solution to deal with the problem of global localization, accumulated location error of odometry, and tracking failure.