Towards Visual Referencing for Location Based Services in Industrial Settings

Conference: ISR 2020 - 52th International Symposium on Robotics
12/09/2020 - 12/10/2020 at online

Proceedings: ISR 2020

Pages: 8Language: englishTyp: PDF

Authors:
Schoepflin, Daniel; Brand, Michael; Gomse, Martin; Schueppstuhl, Thorsten (Institute for Aircraft Production Technology, Hamburg University of Technology, Germany)

Abstract:
With an increasing pursue of automation, the requirements on process quality of Location-Based Services (LBS) increase as well. We consider an Autonomous Mobile Robot (AMR) performing LBS with a carriage unit on the plant premises of an aircraft manufacturer. Due to the manual interaction with the carriage units, location and orientation of the units are considered chaotic, and global localization of the units results in high error margins. An approaching AMR therefore needs to reference itself locally. Approaches with LiDAR sensor solutions allow accurate referencing, but usually do not provide uniqueness or unambiguity and their application to half-automated processes is only limited. A referencing system based on marker tracking provides these key features. However, both the limits and possibilities of such a system are not satisfactory discussed in recent works. We introduce a cascading referencing system based on a unique object ID and marker ID. With that we can verify the objective-task and extract an unambiguous pose and thus, consider this as a viable alternative to feature based localization approaches. Through testing of an implemented solution, we aim to display the limits and possibilities of this approach. Further, we discuss the applicability of this system in settings that use predominantly feature based approaches. In comparison we deem our system relevant for LBS based on three advantages it offers: (1) low-budget implementation, (2) concatenated accuracy leading to an increased scope for AMR path planing, and (3) unambiguity of referencing.