外文翻譯---多模態(tài)六自由度力視覺傳感器融合的姿態(tài)跟蹤

1、附錄 AMulti-modal Force/Vision Sensor Fusion in 6-DOF Pose TrackingAbstract—Sensor based robot control allows manipulation in dynamic and uncertain environments. Vision can be used to estimate 6-DOF pose of an object by mo

2、del-based poseestimation methods, but the estimate is not accurate in all degrees of freedom. Force offers a complementary sensor modality allowing accurate measurements of local object shape when the tooltip is in conta

3、ct with the object. As force and vision are fundamentally different sensor modalities, they cannot be fused directly.We present a method which fuses force and visual measurements using positional information of the end-e

4、ffector. By transforming the position of the tooltip and the camera to a same coordinate frame and modeling the uncertainties of the visual measurement, the sensors can be fused together in an Extended Kalman filter. Exp

5、erimental results show greatly improved pose estimates when the sensor fusion is used.I. INTRODUCTIONRobot control in unstructured environments is a challenging problem. Simple position based control is not adequate,if t

6、he position of the workpiece is unknown during manipulation as uncertainties present in robot task prevent the robot from following a preprogrammed trajectory. Sensor based manipulation allows a robot to adapt to a dynam

7、ic and uncertain environment. With sensors the uncertainties of the environment can be modeled and the robot can take actions based on the sensory input. In visual servoing the robot is controlled based on the sensory in

8、put from a visual sensor. A 3-D model of the workpiece can be created and 6-DOF pose of the object can be determined by pose estimation algorithms. Visual servoing enables such tasks as tracking a moving object with an e

9、nd-effector mounted camera. However, a single camera visual measurement is often not accurate in all degrees of freedom. Only the object translations perpendicular to the camera axis can be determined synchronization of

10、the positional information and visual measurement. Otherwise vision will give erroneous information while the end-effector is in motion.In this paper, we present how vision and force can be fused together taking into acc

11、ount the uncertainty of each individual measurement. A model based pose estimation algorithm is used to extract the unknown pose of a moving target. The uncertainty of the pose depends on the uncertainty of the measured

12、feature points in image plane and this uncertainty is projected into Cartesian space. A tooltip measurement is used to probe the local shape of the object by moving on the object surface and keeping a constant contact fo

13、rce. An Extended Kalman filter (EKF) is then used to fuse the measurements over time by taking into account the uncertainty of each individual measurement. To our knowledge this is the first work using contact informatio

14、n to compensate the uncertainty of visual tracking while the tooltip is sliding on the object surface.II. RELATED WORKReduction of measurement errors and fusion of several sensory modalities using a Kalman filter (KF) fr

15、amework is widely used in robotics, for example, in 6-DOF pose tracking [1]. However, in visual servoing context Kalman filters are typically used only for filtering uncertain visual measurements and do not take into acc

16、ount the positional information of the end-effector. Wilson et al. [2] proposed to solve the pose estimation problem for position-based visual servoing using the KF framework as this will balance the effect of measuremen

17、t uncertainties. Lippiello et al. propose a method for combining visual information from several cameras and the pose of the end effector together in KF [3]. However, in their approaches the KF can be understood as a sin

18、gle iteration of an iterative Gauss-Newton procedure for pose estimation, and as such is not likely to give optimal results for the non-linear pose estimation problem.Control and observation are dual problems. Combining

19、of force and vision is often done on the level of control [4], [5], [6]. As there is no common representation for the two sensor modalities combining the information in one observation model is not straight forward. Prev

20、ious work on combining haptic information with vision in observation level primarily uses the two sensors separately. Vision is used to generate a 3D model of an object and a force sensor to extract physical properties s