Robocentric Mapping and Localization in Modified Spherical Coordinates with Bearing Measurements

In this paper, a new approach to robotic mapping is presented that uses modified spherical coordinates in a robotcentered reference frame and a bearing-only measurement model. The algorithm provided in this paper permits robust delayfree state initialization and is computationally more efficient than the current standard in bearing-only (delay-free initialized) simultaneous localization and mapping (SLAM). Importantly, we provide a detailed nonlinear observability analysis which shows the system is generally observable. We also analyze the error convergence of the filter using stochastic stability analysis. We provide an explicit bound on the asymptotic mean state estimation error. A comparison of the performance of this filter is also made against a standard world-centric SLAM algorithm in a simulated environment.