Filtering and Corner Detection in Predictive Robotic Contour Following

This paper shows how path filtering and corner detection improves robotic look-ahead vision-based contour following using laser-camera-triangulation. The work is motivated by contour following tasks such as sewing, grinding or the application of adhesives with an industrial robot, where position and orientation of the robot’s tool have to be aligned to the contour. Common applications use either manual teach-in mechanisms for the robot path or offline generated paths. Especially in small series production the manual teach-in is time consuming and thus cost-intensive. Therefore the aim is to perform the contour following more autonomously, which requires reliable path filtering and approximation as well as robust corner detection. We describe different mechanisms to calculate path characteristics such as orientations along the contour for 6D contour following. We compare linear and quadratic regression in R(exp 3) and show the reduction of noise and contour errors. Moreover we present an intuitive approach for corner detection in R(exp 3) that is built upon a finite state machine. Keywords: Industrial Robot, Contour Following, Visual Control Conference Topic: Industrial Robots