M. Jagersand, O. Fuentes, and R. Nelson. Experimental evaluation of uncalibrated visual servoing for precision manipulation. In Int. Conf. on Robotics and Automation, 1997.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....estimate of the image Jacobian or interaction matrix in order to compute the control law and this usually depends in part on the parameters of the image feature as well as the unknown depth z. This is a common feature of all IBVS algorithms and a number of methods exist to overcome the di#culties [10, 12, 15, 16]. 3 The dynamics of the optical axis in closed loop IBVS In this section a comparative theoretical analysis of a range of characteristic image feature errors that have been proposed in recent literature is undertaken. We focus on understanding the dynamic response of the closed loop IBVS control ....

M. Jagersand, O. Fuentes, R. Nelson, "Experimental evaluation of uncalibrated visual servoing for precision manipulation", Proc ICRA'97, Vol. 3, pp. 28742880, Albuquerque, April 1997.

....cameras and conclude that multiple cameras can mean greater imagebased errors, depending on the quality criteria used, e.g. 3D or image based. Basic robot knowledge is used to derive 3 Cartesian motions and a Jacobian is produced. A trust region is developed for movement based on the Jacobian in [9], the correct adaptation of the Jacobian for uncertainties is required for significant movement in the arm and image space. 10] successfully use the Broyden method for calculating the Jacobian and target following; the method is restricted by the need to calculate good initial conditions. Many ....

....In Section 4, an exhaustive set of points for goal and initial points are examined to see the effectiveness of the test motion method for different control step sizes. TheuseofafixedWeighting is not efficient for all motions over the arm joint space and the calculation of the trust region method [9] is too complex for real time implementation. A heuristic function to vary the gain is derived from the magnitude of the error from the present position error (err) and the screen width (screen)inpixels: 2.2 Control for Observing Target and Gripper There are many occasions when an object moves ....

Jagersand M et al., "Experimental evaluation of uncalibrated visual servoing for precision manipulation", IEEE Proc ICRA, 1997, pp2874-2880.

....the 3D shape and dimensions of the observed object is still necessary to determine the desired value of the same 3D parameters. Another method in image based visual servoing consists in numerically estimating the coecients of the interaction matrix, without taking into account its analytical form [15, 19]. Contrarily to the previous ones, this method does not need any 3D a priori knowledge. However, it is unfortunately impossible to demonstrate and to ensure its stability. In this paper, we present how the 2 1 2 D visual servoing scheme we have recently developed [24, 25] can be used with ....

M. Jgersand, O. Fuentes and R. Nelson. Experimental evaluation of uncalibrated visual servoing for precision manipulation. IEEE Int. Conf. on Robotics and Automation, Vol. 3, pp. 2874-2880, Albuquerque, New Mexico, April 1997.

....exact same way, i.e. in which x 1 = x 2 . Let us also suppose that the measurement accuracy is oe 1 = oe 2 = 0:2. Then, if we take the constraint into consideration, we can decrease the standard deviation p 2 times. 2.3. 8 Implementation The papers (Nelson et al. 1995) Fuentes et al. 1996) (Jagersand et al. 1997), Kreinovich et al. 1998) describe the actual implementation of several virtual tools for grasping and manipulation with the Utah MIT hand. This implementation does lead to an improved telemanipulation performance. 3 INTEGRATING STATISTICAL DOMAIN KNOWLEDGE WITH DATA Often, instead of crisp ....

M. Jagersand, O. Fuentes and R. C. Nelson, "Experimental Evaluation of Uncalibrated Visual Servoing for Precision Manipulation", Proceedings of the 1997 IEEE International Conference on Robotics and Automation, Albuquerque, New Mexico, April 1997.

....is camera supported fine motion control. Affine Visual Servoing [4] may be applied to such tasks. The changes in the shape of image contours are the input of a feed forward controller. In recent years the idea of using uncalibrated cameras for visual guidance has found increasing interest (c.f. [14, 5]) However, nearly all require some initial and subsequent to perform well. CMAC neural networks may tackle the problem of dimensionality; in [8] 12 inputs represent four joint positions of the robot, four image parameters and their desired changes. The outputs are the control signals for the ....

M. J agersand, O. Fuentes, and R. Nelson. Experimental evaluation of uncalibrated visual servoing for precision manipulation. In Proc. IEEE Int. Conf. Robot. Automat., pages 2874--, 1997.

....switched off, this same robotic arm acts as a simplified virtual tool . Different movement types transform the robotic arm into different virtual tools. The actual implementation of this virtual tool idea for the Utah MIT hand is described in (Nelson et al. 1995) Fuentes et al. 1996) (Jagersand et al. 1997), Kreinovich et al. 1998) This implementation does lead to an improved telemanipulation performance. 3 TOWARDS INTEGRATING FUZZY DOMAIN KNOWLEDGE WITH DATA 3.1 THE NEED FOR FUZZY DOMAIN KNOWLEDGE Often, experts can only formulate their knowledge in terms of words from natural language. For ....

M. Jagersand, O. Fuentes and R. C. Nelson, "Experimental Evaluation of Uncalibrated Visual Serving for Precision Manipulation", Proceedings of the 1997 IEEE International Conference on Robotics and Automation, Albuquerque, New Mexico, April 1997.

....and Allen, 1996) demonstrated 2D alignment of an eye on hand manipulator using rotational epipolar motion. Both (Hager et al. 1995) and (Cipolla and Hollinghurst, 1994) exploit a nearly uncalibrated stereo camera setup. Recent work shows the feasibility using the well known image Jacobian. In (Jagersand et al. 1997) the usefulness of adaptive differential feedback employing a visual motor Jacobian was shown. In (Sutanto et al. 1997) the idea of exploratory movements for dynamic image Jacobian estimation was demonstrated. This work describes a different way of uncalibrated hand eye coordination. It is based ....

Jagersand, M., O. Fuentes and R. Nelson (1997). Experimental evaluation of uncalibrated visual servoing for precision manipulation. In: Proc. IEEE Int. Conf. Robot. Automat.. pp. 2874--.

....epipolar motion. Both Hager [4] and Hollinghurst Cipolla [2] exploits a nearly uncalibrated stereo camera setup. Also Hager introduced the ideas of projective invariance into the field of uncalibrated visual control [3] Recent work shows the feasibility using the well known image Jacobian. In [6] the usefulness of adaptive differential feedback employing a visual motor Jacobian was shown. In [10] the idea of exploratory The work described in this paper has been funded by the German Research Foundation (DFG) in the project SFB 360. movements for dynamic image Jacobian estimation was ....

M. Jagersand, O. Fuentes, and R. Nelson. Experimental evaluation of uncalibrated visual servoing for precision manipulation. In Proc. IEEE Int. Conf. Robot. Automat., pages 2874--, 1997.

....camera supported fine motion control. Affine Visual Servoing [4] may be applied to such tasks. The changes in the shape of image contours are the input of a feed forward controller. In recent years the idea of using uncalibrated cameras for visual guidance has been found increasing interest (c.f. [15, 16, 5, 14]) However, nearly all require some initial and subsequent to perform well. c) CMAC neural networks may tackle the problem of dimensionality; in [10] 12 inputs represent four joint positions of the robot, four image parameters and their desired changes. The outputs are the control signals for the ....

M. J agersand, O. Fuentes, and R. Nelson. Experimental evaluation of uncalibrated visual servoing for precision manipulation. In Proc. IEEE Int. Conf. Robot. Automat., pages 2874--, 1997.

....requirement can often be satisfied even when camera and actuator models are themselves imprecisely known 1 . As a result, it is possible to position a robot with high accuracy without relying on an accurate hand eye calibration. This property, which has been observed in several special cases [44, 23, 16, 14, 27, 25, 15, 26, 41], follows from the fact that often both the position and orientation of the robot in its workspace and the features defining a set of desired poses can be simultaneously observed through the same camera system. This observation has motivated a great deal of research on solving the algorithm design ....

....robot in its workspace and the features defining a set of desired poses can be simultaneously observed through the same camera system. This observation has motivated a great deal of research on solving the algorithm design problem for imprecisely calibrated or uncalibrated camera actuator systems [26, 5, 45, 23, 24], and related research on choosing feature configurations to maximize accuracy in the presence of noise [32, 40] However, little is known about general conditions under which the accomplishment of a given task can be determined in the presence of camera model uncertainty, even though this is a ....

M. Jagersand, O. Fuentes, and R. Nelson. Experimental evaluation of uncalibrated visual servoing for precision manipulation. In Proc., ICRA, pages 2874--2880, 1997.

....of the task function T under projective transformations. Theorem 9.3, the main result of Chapter 9, states that a given task is decidable on a weakly calibrated two camera class if and only if the task is a projective invariant. This result thus serves to underscore the observation made in [90, 97, 85, 98] that accurate metric information is not needed for the accomplishment of many types of positioning tasks with a stereo vision system. The two camera models considered in this chapter are pairs of projective camera models that map subsets of P 3 containing V , into P 2 Theta P 2 . ....

....of the task function T under projective transformations. The main result of this chapter, Theorem 9.3, states that a given task is decidable on a weakly calibrated two camera class if and only if the task is a projective invariant. This result thus serves to underscore the observation made in [90, 97, 85, 98] that accurate metric information is not needed for the accomplishment of many types of positioning tasks with a stereo vision system. The two camera models considered in this chapter are pairs of projective camera models which map subsets of P 3 containing V , into P 2 Theta P 2 . ....

M. Jagersand, O. Fuentes, and R. Nelson. Experimental evaluation of uncalibrated visual servoing for precision manipulation. In Proc. IEEE Int. Conf. on Robot. and Automat., pp. 2874--2880, 1997. BIBLIOGRAPHY 121

....requirement can often be satisfied even when camera and actuator models are themselves imprecisely known 1 . As a result, it is possible to position a robot with high accuracy without relying on an accurate hand eye calibration. This property, which has been observed in several special cases [44, 23, 16, 14, 27, 25, 15, 26, 41], follows from the fact that often both the position and orientation of the robot in its workspace and the features defining a set of desired poses can be simultaneously observed through the same camera system. This observation has motivated a great deal of research on solving the algorithm design ....

....robot in its workspace and the features defining a set of desired poses can be simultaneously observed through the same camera system. This observation has motivated a great deal of research on solving the algorithm design problem for imprecisely calibrated or uncalibrated camera actuator systems [26, 4, 45, 23, 24], and related research on choosing feature configurations to maximize accuracy in the presence of noise [32, 40] However, little is known about general conditions under which the accomplishment of a given task can be determined in the presence of camera model uncertainty, even though this is a ....

M. Jagersand, O. Fuentes, and R. Nelson. Experimental evaluation of uncalibrated visual servoing for precision manipulation. In Proc., ICRA, pages 2874--2880, 1997.

....with endeffector mounted camera the desired position. For vision based sensors features are extracted from an image. Then the corresponding positions are calculated using the known robot and sensor characteristics. Unknown kinematics require learning of the hand eye transformation (see e.g. [vdSG97, JFN97]) In industrial applications this is not a problem since the kinematic parameters are known and the deviations between the tool center point and the sensed desired position will be small. So step 1 is a standard problem which will not be outlined here (see e.g. HHC96] Step 2 requires the ....

M. Jagersand, O. Fuentes, and R. Nelson. Experimental evaluation of uncalibrated visual servoing for precision manipulation. In Proc. IEEE Int. Conference on Robotics and Automation, Albuquerque, New Mexico, April 1997.

....this pictorially. In addition, since the vision system provides continuous pose information, the introduction of differential visual error feedback allows precise positioning despite slight inaccuracies or simplifying approximations in the model. Several flavors of visual servoing exist, e.g. [2, 3, 5, 6, 7, 10, 12, 14, 15, 18, 19, 20] 2 ) Visual models suitable for specifying simple visual alignments have also been studied [21, 11, 9, 8] Learning by Watching. Model based work in visionbased teaching of a robot, as pioneered by Kuniyoshi, Seeing, Understanding and Doing Human Task [24] and Ikeuchi ( Assembly Plan From ....

....described in object based coordinate frames, and during a task each time an object is grasped or regrasped the manipulation model changes. The general principle is for each step in a manipulation task to use the simplest possible (lowest DOF) servoing movement. From our experimental evaluation in [10, 15], we know that the smaller and simpler the manipulation model is, the more robust model estimation and servoing is achieved. Also controlling fewer DOF s require fewer visual measurements, and a smaller task description. The three main classes of servoing movements are transportation, alignment ....

Jagersand M. Nelson R. Fuentes O. "Experimental Evaluation of Uncalibrated Visual Servoing for Precision Manipulation " In Proc. of Int. Conf. on Robotics and Automation 1997.

.... has also proliferated [2, 4, 10, 12, 27, 28, 32, 34, 39, 43] Some of this work considers task composition and discrete event systems to link active vision modules, e.g. 29, 30, 14] In our own work, we have begun the development of modular systems for hand eye coordination and visual tracking [19, 21, 25, 26]. Our work differs from other work in several ways. First, we emphasize a modular approach where task primitives are defined for simple subtasks that occur repeatedly in the domain. These primitives are composed to perform complex task via consistent design principles, where composition can occur ....

....is to use the simplest possible servoing movement which suffices for each step in a manipulation task. In other words, we use the set of primitives which accomplish the task with contraints on the fewest degrees of freedom (DOF) Our experimental evaluations of control level visual servoing in [26] show that smaller and simpler manipulation models result in more robust model estimation and servoing. Also, controlling fewer DOF requires fewer visual measurements and a smaller task description. Simpler and more robust tracking can be used for coarse movements, and sensitive, computationally ....

M. Jagersand, O. Fuentes, and R. Nelson. Experimental evaluation of uncalibrated visual servoing for precision manipulation. In Proc., ICRA, pages 2874--2880, 1997.

....observing both controlled features (e.g. those on a robot manipulator) and a (static) target. It has been observed that there exist approaches which can achieve precise task positioning, even though the relationship between the manipulator and the observing cameras may not be precisely known. [11, 8, 12]. In particular, prior work has shown that one route toward defining tasks performable with absolute precision is to take advantage of the invariance of certain image level constructions to perspective projection ( 7, 9, 2] It is also known that there are tasks which, under certain conditions, ....

M. Jagersand, O. Fuentes, and R. Nelson. Experimental evaluation of uncalibrated visual servoing for precision manipulation. In Proc., ICRA, pages 2874-- 2880, 1997.

....to set the possible choices for c J s . In fact, three different cases have been considered in the literature: ffl c J s = c J s (t) In that case, the image Jacobian is numerically estimated during the camera motion without taking into account the analytical form given by (2) [13,15] (neural networks have also sometimes been used [27] This approach seems to be very interesting if any camera and robot models are available. However, it is impossible in that case to demonstrate when condition (4) is ensured. Furthermore, initial coarse estimation of the image Jacobian may ....

M. Jagersand, O. Fuentes, and R. Nelson. Experimental evaluation of uncalibrated visual servoing for precision manipulation. In IEEE Int. Conf. on Robotics and Automation, volume 3, pages 2874--2880, Albuquerque, New Mexico, April 1997.

....by [14] whose findings suggest that for a weakly calibrated stereo vision system, there ought to be a close relationship between the decidability of a given task T (f) 0 and the invariant properties of T under projective transformations. This result also underscores the observation made in [10, 15, 16, 17] that accurate metric information is not needed for the accomplishment of many types of positioning tasks with a stereo vision system. This paper continues a line of research started in [18] and subsequently pursued in [19, 20, 21] Notation: Throughout this paper, prime denotes matrix transpose, ....

M. Jagersand, O. Fuentes, and R. Nelson. Experimental evaluation of uncalibrated visual servoing for precision manipulation. In Proc., ICRA, pp. 2874-- 2880, 1997.

No context found.

M. Jagersand, O. Fuentes, and R. Nelson. Experimental evaluation of uncalibrated visual servoing for precision manipulation. In Int. Conf. on Robotics and Automation, 1997.

No context found.

Martin Jagersand, Olac Fuentes and Randal C. Nelson, "Experimental Evaluation of Uncalibrated Visual Servoing for Precision Manipulation", Proceedings of the 1997 IEEE International Conference on Robotics and Automation, Albuquerque, New Mexico, April 1997.

No context found.

M. Jgersand, O. Fuentes and R. Nelson. Experimental evaluation of uncalibrated visual servoing for precision manipulation. IEEE Int. Conf. on Robotics and Automation, Vol. 3, pp. 2874-2880, Albuquerque, New Mexico, April 1997.

No context found.

M. Jagersand, O. Fuentes, R. Nelson, "Experimental evaluation of uncalibrated visual servoing for precision manipulation", Proc ICRA'97, Vol. 3, pp. 28742880, Albuquerque, April 1997.

No context found.

M. Jgersand, O. Fuentes, R. Nelson, "Experimental evaluation of uncalibrated visual servoing for precision manipulation", IEEE Int. Conf. on Robotics and Automation, ICRA'97, Vol. 3, p. 2874-2880, Albuquerque, New Mexico, April 1997.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC