Nonlinear Visual Mapping Model for 3-D Visual Tracking With Uncalibrated Eye-in-Hand Robotic System

Abstract

Abstract-A new control scheme for uncalibrated robotic visual tracking problem is proposed that compromises the computational expenses of overall system with offline modeling and online control. A nonlinear visual mapping model for the uncalibrated hand-eye coordination is first proposed with an artificial neural network implementation. An online visual tracking controller is then developed together with a real-time motion planner. To improve the system performance, the control scheme is also integrated with a feedforward controller to compensate unknown object motions. Extensive simulations and experiments demonstrate the effectiveness of the proposed control scheme.Index Terms-Hand/eye coordination, image Jacobian matrix, neural networks, nonlinear mapping.

Task-Independent Robotic Uncalibrated Hand-Eye Coordination Based on the Extended State Observer

“…x 2 (t); z 3 (t) ! x 3 (t): (8) According to the definition in (2), x 3 (t) is the summation of the unknown function and the external disturbance in system (1). Thus, although f (x; _ x; t) and w(t) in (1) are unknown, the extended state z 3 in (4) can still have real-time estimation for their summation x 3 (t).…”

“…One is the image Jacobian matrix-based approach [1], [2], in which the unknown static and nonlinear relations between the robot (hand) and visual sensors (eye) are approximated by the linear image Jacobian matrix that is dynamic and should be estimated iteratively online. The other is the artificial neural network (ANN)-based approach, in which the hand-eye relations are modeled by a neural network [6], and/or the coordination controllers are realized in a neural network way [8].…”

“…In the ANN-based approaches, the hand-eye relationship model is trained off-line, no matter whether the model is explicitly described [7] or not [8], and used as a part of online control. However, since the off-line training of the ANN needs to gather a large set of training samples, feasibility and applicability of this approach is restricted in practice.…”

Calibration-free robotic eye-hand coordination based on an auto disturbance-rejection controller

“…. ; n (15) characterize the system's dynamic performance. We take advantage of a nonlinear combination of the errors to realize the control [23] u 0 = b 1 fal(e 1 ; ; ) + .…”

“…Since the image Jacobian matrix describes a temporary linear relationship, the capacity of the neural network to approach a nonlinear function is not sufficiently exploited. In [15], a nonlinear visual mapping model is invented to describe the dynamic relationship between the system error observed by a camera and the system control. The artificial neural network is used to map the nonlinear model and produce control accordingly, which improves the system performance to a great extent.…”