Robotic Space Exploration and Navigation in Low Visibility Conditions

Abstract

The dream of space exploration became attainable with the advanced devel- opments of autonomous technologies. NASA triggered a paradigm shift in space exploration by landing the Mars helicopter, Ingenuity, and the Perseverance Rover for collaboratively exploring the Martian surface. The Ingenuity scouts terrain in- formation which can be utilized to select safe traversable paths for the Perseverance Rover to follow. Nevertheless, relative localization remains of fundamental impor- tance for the rover to track the selected safe trajectories. In this work, a relative localization system is developed relying on neuromorphic vision-based measurements (NVBM) from a neuromorphic vision sensor (NVS) and inertial measurements from an inertial measurement unit (IMU). The measurement streams are fused in pro- posed state estimators, landmark tracking Kalman filter (LTKF) and translation decoupled Kalman filter (TDKF) for tracking and relative localization, respectively. Neuromorphic vision marked a new paradigm in vision-based measurements where the NVS pixels asynchronously generate events activated by light intensity varia- tions. The major drawback of neuromorphic vision is the loss of NVBM in static scenarios due to illumination invariance. Therefore, high frequency active landmarks are adopted as salient features for consistent event firing to facilitate reconstructing the relative pose between the Ingenuity and Perseverance Rover. A novel event- based landmark identification algorithm using Gaussian Mixture Models (GMM) is proposed for matching the landmarks detected 2D image points to their 3D world correspondences for relative localization. Consequently, the space agents poses are estimated by the TDKF. The proposed system has been tested in a variety of ex- periments where an NVS and IMU are mounted on an unmanned aerial vehicle (UAV) resembling Ingenuity and flickering landmarks attached to a ground vehicle representing the rover to simulate the targeted space scenario. The proposed system demonstrates a positioning error of 5 mm within 7 meter range.

Date of Award	Jul 2022
Original language	American English