Journal of Field Robotics · 2007 · 118 citations · 23 references
ComputingAutonomous SystemsRobotics EngineeringAerial RoboticsComputer VisionSensor FusionAutonomous NavigationLocalization UncertaintyMappingPose EstimationRobot LocalizationSimultaneous LocalizationExtended Kalman FilterExtended Kalman Filter LocalizationInertial SensingLocalization RobustnessSpatial Uncertainty AnalysisSpatial Uncertainty QuantificationSpatial Uncertainty RepresentationUnmanned Aerial Vehicle Navigation
The paper develops a robust inertial sensor–based SLAM algorithm for UAVs that uses only bearing observations. The system uses a single color camera to extract terrain features, estimates the UAV’s 6‑DOF pose and 3‑D feature positions with an EKF that delays initialization, and associates features via a multihypothesis innovation gate based on spatial uncertainty. Offline tests on flight‑test data show the algorithm can estimate UAV pose and map features. © 2007 Wiley Periodicals, Inc.
Abstract This paper presents the on‐going design and implementation of a robust inertial sensor based simultaneous localization and mapping (SLAM) algorithm for an unmanned aerial vehicle (UAV) using bearing‐only observations. A single color vision camera is used to observe the terrain from which image points corresponding to features in the environment are extracted. The SLAM algorithm estimates the complete six degrees‐of‐freedom motion of the UAV along with the three‐dimensional position of the features in the environment. An extended Kalman filter approach is used where a technique of delayed initialization is performed to initialize the three‐dimensional position of features from bearing‐only observations. Data association is achieved using a multihypothesis innovation gate based on the spatial uncertainty of each feature. Results are presented by running the algorithm off‐line using inertial sensor and vision data collected during a flight test of a UAV. © 2007 Wiley Periodicals, Inc.
23