Enhancing Accuracy of Estimating Pedestrian Velocity and Walking Distance in the Workplace with Pose Graph Optimization

Indoor positioning systems are crucial for workplace analysis in GNSS-denied environments, offering insights into individual workloads through accurate localization and movement tracking. This paper addresses challenges in estimating pedestrian velocity and walking distance using a hybrid approach combining Pedestrian Dead Reckoning (PDR) with Bluetooth Low Energy (BLE) and pose graph optimization. The study introduces a methodology integrating PDR for relative positioning, BLE for absolute positioning via particle filtering, and subsequent pose graph optimization to enhance the accuracy of velocity and distance walked. PDR utilizes IMU data to estimate velocity and attitude changes, while BLE RSSI signals are integrated using particle filters with Friis’ law and occupancy-grid-based weighting. The resulting particle filter output serves as the initial state for pose graph optimization, mitigating local minima and discontinuities in location estimates. Evaluation conducted on an open dataset from the xDR Challenge 2023 demonstrates the method’s efficacy. Results indicate that while pose graph optimization does not uniformly improve positional accuracy, it effectively reduces discontinuities inherent in particle filter outputs. PDR-only approaches lack absolute velocity context, resulting in persistent errors in velocity and walking distance estimation. This research contributes to advancing indoor positioning technologies by proposing a robust methodology for accurately estimating velocity and distance walked. By enhancing the reliability of these estimates, the proposed approach facilitates quantitative evaluation of physical demands in diverse workplace environments.