A Real-Time Factor-Graph-Optimized Pedestrian Navigation Method

Inertial measurement units (IMUs) are widely used in Internet of Things (IoT) applications to determine precise self-location for humans. However, biases in accelerometers and gyroscopes can cause significant accumulative positioning errors. Additionally, traditional methods employing the Kalman filter introduce nonlinear errors, which further exacerbate positioning inaccuracies. In this article, we propose a real-time pedestrian navigation method that leverages factor-graph optimization to address these issues. The factor-graph framework is capable of handling nonlinear errors introduced by traditional filter-based approaches and enhances biases estimation accuracy by utilizing more historical data. Moreover, a historical-data-based single zero velocity point detection method is proposed to find a point that is physically closer to zero velocity over a relatively long period. This method provides a more robust and accurate zero-velocity measurement in complex gaits through a long-term judgment. Furthermore, lower uncertainty in zero velocity allows for more precise estimation of IMU biases, thereby improving positioning accuracy. Experimental results demonstrate that the proposed detection method accurately detects zero-velocity points under more complex gaits. In addition, the position error of the proposed optimization-based method is reduced by approximately 80% compared to the filter-based method for low-cost IMUs. These results indicate significant potential for practical applications.