Abstract

This paper describes a Sliding Window Filter (SWF) that is a constant-time approximation to the feature-based batch non-linear least squares Simultaneous Localization and Mapping (SLAM) problem. In particular, we are interested in improving the range resolution of stereo vision for Entry, Descent and Landing (EDL) missions to Mars and other planetary bodies. The goal is to create accurate and precise 3D planetary surface structure estimates by filtering sequences of stereo images taken from an autonomous landing vehicle. More generally we are interested in fast, optimal, relative spatial estimation for mobile robots. The SWF is useful in this context because it can scale from the offline, optimal batch least squares solution to fast online incremental solutions. For instance, if the window encompasses all time, the solution is algebraically equivalent to full SLAM; if only one time step is maintained, the solution is algebraically equivalent to the Extended Kalman Filter SLAM solution; if robot poses and environment landmarks are slowly marginalized out over time such that the state vector ceases to grow, then the filter becomes constant time, like Visual Odometry. Further, the sliding window method exhibits other interesting properties, like reversible data association, out-of sequence measurement updates, and robust estimation across multiple timesteps. We test the SWF with image data captured to emulate EDL conditions for a Mars lander. Experiments show that structure estimates derived from the SWF converge to the optimal result predicted by theory. To the best of our knowledge, this is the first work to show optimal

Citations