LAUSR

Abstract Reliable real-time motion tracking is a crucial element for the proper function of autonomous robotic systems. Unstable environments, for instance such that demonstrate varying light conditions and moving shadows, cause problems for many of the commonly used vision-based position feedback systems. To eliminate the dependency on light, positioning systems based on Radio Frequency Identification (RFID) can be used and are gaining both academic and commercial attention. RFID-based positioning systems can rely on signal strength indicators to calculate the relative distance between RFID tags and the reader. However, they may entail drawbacks due to noise or signal loss caused by unavoidable sources of disturbance. This paper proposes a filter for the preprocessing before the calculation of the position to overcome the problem of blind spots. The paper describes an RFID-based positioning system for an experimental pipe-less plant. The position of each Automated Guided Vehicle (AGV) in the setup is calculated based on a trilateration algorithm using Received Signal Strength Indicators (RSSI). Ambiguous RSSI values that correspond to multiple possible distances between the reader and a tag are reinterpreted based on the probable location in the overall grid of RFID tags. The proposed filter improves the positioning precision from an average error of about 45 millimeters down to less than 25 millimeters by eliminating invalid data points.