The next-best-view for workpiece localization in robot workspace

Workpiece localization is the process of obtaining the location of a workpiece in a reference frame of a robotic workspace. The location (position and orientation) is represented by the transformation between a local frame associated with the workpiece and the specified reference frame in the workspace. In this work, we study the workpiece localization problem without the two commonly adopted restrictive assumptions: the data used to calculate the transformation is readily available and the correspondence between the data sets used for calculation is known. The goal is to automate the localization process starting from efficient data collection to determining the workpiece location in the workspace. We describe a strategy that includes the following aspects: predicting the correspondence between the measured data and the workpiece CAD model data; generating representative vectors that would aid in determining the next-best-view for collecting new information of the workpiece location; evaluating a search region to find the next sensor location that satisfies both the robot kinematics as well as sensor field-of-view constraints while giving the maximum view gain; and calculating the rigid body transformation from the local frame to the world frame to localize the workpiece. Numerical simulation and experimental results are presented and discussed for the proposed strategy.