Robust edge-tracking for industrial stitching of carbon fibre mats using a laser range scanner

Abstract

Nowadays, robust and light-weight parts used in the automobile and aeronautics industry are made of carbon fibres. To increase the mechanical toughness of the parts, the carbon fibres are stitched in the pre-forming process using a sewing robot. However, current systems miss high flexibility and rely on manual programming of each part. The main target of this work is to develop an automatic system that autonomously sets the structure strengthening seams. Therefore, a rapid and flexible following of the carbon textile edges is required. Due to the black and reflective carbon fibres, a laser-stripe sensor is necessary and the processing of the range data is a challenging task. The captured profiles have a high noise ratio and many unwanted fragments such as single carbon fibres and sewing wires towering above the edge. To improve the edge tracking process, three different edge detection methods were used and merged. The first one is a global method called model fit: This method attempts to fit a model of the edge-shape into the profile data. The second global method is called gradient accumulation and this approach uses increasing filter kernels and sums up the gradients. The third method is a local method, which means that it uses just parts of the profile data for calculating the edge. This method is called local weighted voting and calculates a value for every possible edge jump with several weighting factors. All three methods have their strengths and weaknesses and can find the correct edge with a hit probability of approximately 80\%. Due to the fact that the edge has to be found in bulged and tilted profiles, a preprocessing step is added. This preprocessing step filters and flattens the incoming profile data. To increase the edge tracking robustness and the hit probability, the results of the three methods are merged and an additional edge prediction step calculates the final edge with a RANSAC line-fit algorithm. The framework is capable to communicate with the laser sensor over UDP and to receive the profile data, process the data and send the edge position as XML over TCP to the robot controller, which uses the data to correct the robot's path for stitching in real-time. The experimental results demonstrate the feasibility of a fully automated, sensor-guided robotic sewing process. Summing up, this thesis proposes a real time approach where different edge detection methodologies are merged to increase the edge tracking robustness and lift the hit probability to approximately 97\%.<br />