Advancing toward Automated Material Counting and Localization Using RFID, a Boston Dynamics Spot Robot, and a Bayesian Filter Algorithm

The purpose of this research was the deployment and experimental testing of an autonomous terrestrial robot (Boston Dynamics Spot) for radio-frequency identification (RFID)-based probabilistic localization of construction inventory within a mockup laydown yard. The study involved designing and deploying a customized RFID payload comprising Spot overlaid with depth and tracking cameras, an RFID reader, and an RFID-based probabilistic Bayesian filter of fixed RF transmission power (BFFP) algorithm interfaced with a robot operating system (ROS) and RViz. The system was experimentally evaluated in an open laboratory setting based on its performance to read and localize tags in an unknown space. Once all installed RFID tags were localized using the hybrid system, the localization errors were calculated based on true locations. Specifically, the objectives of the study were: (1) to design and deploy a payload that leverages RFID technology, a recursive probabilistic filter for localization, edge computing, and an autonomous mobile platform for material localization in an unmapped space; (2) to adjust the original recursive probabilistic algorithm and ROS-based RViz interface for this study; (3) to calculate the power output of the RFID scanner to capture all tags while optimizing for localization; (4) to identify best practices and future research directions in automated mapping, robotic navigation, and real-time data collection for autonomous construction inventory management. According to the results, the localization error varied considerably at lower power levels (dBm). Ultimately, this study is a contributory attempt in the push toward incorporating more automation and robotics in the construction processes, especially in the wake of labor shortages and stagnation in productivity levels. Within the scope and limitations, the study answered two of the biggest challenges of autonomous inventory management on construction worksites. First, it answered “Where is my inventory?” through the BFFP-based probabilistic localization. It answered, “How much/many is my inventory?” through successful RFID tag readings.