Proximity operations of CubeSats via sensor fusion of ultra-wideband range measurements with rate gyroscopes, accelerometers and monocular vision

Abstract

A robust pose estimation algorithm based on an extended Kalman filter using measurements from accelerometers, rate gyroscopes, monocular vision and ultra-wideband radar is presented. The sensor fusion and pose estimation algorithm incorporates Mahalonobis distance-based outlier rejection and under-weighting of measurements for robust filter performance in the case of sudden range measurements led by the absence of measurements due to range limitations of radar transceivers. The estimator is further validated through an experimental analysis using low-cost radar, IMU and camera sensors. The pose estimate is utilized to perform proximity operations and docking of Transforming Proximity Operations and Docking Service (TPODS) satellite modules with a fixed target.

Figures (7)

• Figure 1: TPODS module uses UWB radar in two way ranging mode to measure distance to stationary anchors. Rate gyroscope measures angular velocity and accelerometer measurements enable visibility of translation velocity through indirect measurement of frictional forces. Monocular vision data is further utilized when UWB measurements are not reliable near 10 cm of the target.
• Figure 2: Geometry of the TPODS System perspective projection problem. The camera reference frame ($\Hat{C}$) which is on the chaser TPODS module can be related to the target TPODS module ($\Hat{B}$) by relating the respective position vectors to the ith feature.
• Figure 3: The TPODS module uses LED based marker patterns, which are unique for each face, for monocular vision based pose estimation. The monocular camera captures an image frame and converts it in grayscale, which is further used to obtain all the blobs present in the image. The blobs outside certain threshold are rejected and only unique features are used for pose estimation.
• Figure 4: Simulated range measurements are corrupted with measurement noise and outliers. The module is commanded to move under constant acceleration in X and Y directions alternatively. The range measurements are sampled from a Gaussian distribution with a true range as mean and standard deviation of 1 cm. For the outlier generation, $10\%$ of the total samples are randomly chosen and re-sampled from a Gaussian distribution with a standard deviation of 10 cm.
• Figure 5: The DWM1000 UWB radio used in the loco node possesses different gains for different planes of motionsorgel2005influence. This is undesirable and might result in different pose estimation biases while moving in various planes OG_fusion. However, since the module is constrained to move in a single plane, such effects are not of concern for this study. Each anchor node is mounted above $15cm$ from the ground to eliminate any interference of the multi-path reflection from the ground. In addition, the node on the module is mounted on the top to have a clear line of sight with all four anchors. The accompanying video submission includes more details about the setup.