Controlling a Robotic Stereo Camera Under Image Quantization Noise

摘要: In this paper, we address the problem of controlling a mobile stereo camera under image quantization noise. Assuming that pair images set targets is available, moves through sequence Next-Best-Views (NBVs), i.e., views minimize trace targets' cumulative state covariance, constructed using realistic model rig captures noise and Kalman Filter (KF) fuses observation history with new information. The proposed algorithm decomposes control into two stages: first NBV computed in relative coordinates, then to realize view fixed global coordinate frame. This decomposition allows drive pose effectively realizes coordinates while satisfying Field-of-View constraints task particularly challenging complex sensing models. We provide simulations real experiments illustrate ability system accurately localize sets targets. also propose novel data-driven technique characterize unmodeled uncertainty, such as calibration errors, at pixel level show method ensures stability KF.