The "feature CMAC": a neural-network-based vision system for robotic control

Abstract

A strategy for locating and grasping a target object in an unknown position using a robotic manipulator equipped with a CCD camera is described. Low-level trajectory and joint control during the grasping operation is handled by the manipulator S conventional motion contrcrller using target-pose data provided by an artificial-neural-network-based vision system. The Feature CMAC is a self-organizing neural network that eficiently transforms images of a target into estimates of its location and orientation. The approach emulates biological systems in that it begins with simple image features (e.g., comers) and successively combines them to form more complex features in order to determine object position. Knowledge of camera parameters, camera position and object models is not required since that information is incorporated into the network during a training procedure wherein the target is viewed in a series of known poses. The manipulator is used to generate the training images autonomously. No training of connection weights is required; instead, training serves only to define the network topology which requires just one pass through the training images. Experiments validating the effectiveness of the strategy on an industrial robotic workcell are presented.

Tutorial Review of Bio-Inspired Approaches to Robotic Manipulation for Space Debris Salvage

“…CMAC has been applied as lookup tables to reproduce input-output functions defined by the kinematics of a robotic manipulator [80]. CMAC has been applied to the grasping control of a robotic manipulator using CCD camera images transposed to object locations that were passed to a conventional robot controller [81]. The look-up table representation is not consistent with biology however -it appears that motor adaptation does not involve the composition of look-up tables rather than the forming a full and adaptable model which can extrapolate beyond the initial training data [82].…”

Intelligent Control of Robotic Systems

“…Particular attention has been paid to the problem of vi suo-motor coordination [249], [250], [ 196], [235] , [113], [324], [73], [45], [188], [27], [83], in particular for eye-head and arm-eye systems. Sensor-based control is a very efficient method of overcoming the problems with robot model and environment uncertainties, because sensor capabilities help in the adaptation process without explicit control intervention.…”

Coordination of an Uncalibrated 3-D Visuo-Motor System Based on Multiple Self-Organizing Maps

“…Visuo-motor models are either estimated analytically during the execution of the task on-line or learned off-line prior to the execution of the task. Artificial neural techniques can be used to learn the non-linear relationships between features in the images and manipulator joint angles (1), (2) . In Ref.…”

“…In Ref. (2), an approach used to train an uncalibrated industrial robot is proposed. A neural network provides the estimate of the pose of the target in the manipulator coordinate frame.…”