Accuracy and Precision Assessment of Binary Fiducial Marker-Based Sensor Spatial Fusion Frameworks for Manufacturing

In large format additive manufacturing (LFAM), in-situ process monitoring is used for defect detection, process model input, and quality assurance. The accuracy and sensitivity of the sensors, like infrared cameras, are dependent on position and pose relative to their measurement location. Commercial solutions are expensive and leave the sensing environment location-locked, which is not compatible with modular and interactive manufacturing cells, such as collaborative manufacturing robotics. This work assessed error from using binary fiducial markers and visual cameras as a position and pose estimation method in a manufacturing environment for application to LFAM where cost-effective, scalable, and flexible spatial frameworks are required. AruCo Markers and April tags were assessed at close range (sub 2 meters) and long range (greater than 2 meters) with singular and multiple visual cameras to determine single-point and multi-point triangulation accuracy in various optical configurations. The absolute and relative percent errors were determined by using a point-cloud scanner and distances and poses were measured and compared with the pose/position estimation. This work provided a cost-effective recommendation for two length scales to perform sensor pose, position, and velocity estimation, and reports their absolute and relative uncertainty and error, which is needed for accurate and representative sensor values measurements.