Camera Based Concept for Enhancement of IRB Accuracy Using Fixed Features and SLAM

A standard IRB (Industrial Robot) of today is not capable of positioning itself to within better than 2–5mm relative to a given structure or object. In order to be able to perform high precision manufacturing, the requirement is less than 0.5mm. Existing solutions for 6DOF measurements to enhance the accuracy are not only expensive but also lack flexibility. An affordable, accurate concept for IRB closed loop position feedback is presented here. This article proposes a method for using a camera together with fixed features to enhance the accuracy of an IRB in industrial manufacturing processes. The method is based on SLAM (Simultaneous Localization and Mapping), used to structure collected information about the features (mapping), and for the IRB to locate itself among the features (localization). The concept covers a scenario in which a set of features are placed within the working area. A camera is mounted on the IRB to observe (measure) the relative pose of the feature set. When the system is running observation of only one feature increases accuracy. However, in normal operation more features are observed which will increase accuracy further. The system setup is a standard IRB, ABB4400, on which a single camera with IR flash is mounted and used as sensor. The features are made out of reflectors that are placed in the production cell. The camera is also equipped with an IR filter to improve the signal to noise ratio, simplifying image processing and accelerating execution. The execution time is critical since the goal is to sample at such a rate that as to cover the dynamics of the robot.