Modeling, implementation and evaluation of SLAM for determination of anchor configurations

Status: abgeschlossen
Betreuer: Horst Hellbrück , Manfred Constapel
Student: Yuefeng Liu


Industry 4.0 applications become more and more important. Real-time data acquisition allows the construction of digital representation of real production lines. These representations ("digital twins") create great opportunities for process management and optimization. To achieve that, the system needs to be able to accurately pinpoint the location of goods, tools and production parts. Research and practice show that localization systems can be realized with a high degree of accuracy by using Ultra-Wide-Band (UWB) technology. Anchors (transmitters) are placed all over the production line, all moveable objects can be equipped with or become a tag (receiver). Thus, a tag provides its position within the production line using distance measurements from several anchors. For this, positions of all anchors need to be known in advance. The determination of the precise positions of anchors in an unknown production line can be treated as a problem of SLAM (simultaneous localization and mapping).


Consider a randomly placed robot (e.g. Roomba) equipped and connected to a UWB tag, is placed at an unknown position and unknown environment (the production line). The robot will (randomly) walk around the environment, in order to capture the positions of arbitrarily placed UWB anchors via distance measurements. The goal is to simultaneously create a map and estimate the own position and pose with noisy odometry readings. This requires prior knowledge of anchor positions. Due to non-perfect pose estimations, the anchor positions become inaccurate over time. On the one hand, the mutual aim of SLAM is to provide reliable pose estimation, on the other hand accurate anchor localization.


  • Comprehensive and well explained system design
  • Reasonably introduced models for the system dynamics
  • Systematic analysis and proper evaluation of the proposed solution


  • Development and evaluation of a solution to the previously stated SLAM problem
  • Analytical modeling of relevant system dynamics and noise (e.g. motion, odometry readings, distance measurements)
  • Simulation of the solution for a given production line with 44 anchors and an area of about 2000 square meters


  • Good programming skills in Python or an akin programming language
  • Basic knowledge about Bayesian filters (e.g. Kalman filter)
  • Good knowledge of probability theory