Title: MM-Scale Localization of RFID-Based Motion Capture and Localization Systems
Dr. Durgin, Advisor
Dr. Taylor, Co-Advisor
Dr. Y. Zhang, Chair
The objective of the proposed research is to build the first RFID-based motion capture system in 3D motion, e.g., ballerina using M backscatter tags and N readers to capture motion by estimating positions of tags in 3D motion. Nonlinear state estimation is designed and applied to enhance localization accuracy for achieving motion-capture grade localization at various noise levels in a range of several meters. This proposal presents and analyzes the design of nonlinear least-squares estimation (NLE) for state estimation of real-time RFID localization and tracking in 2D motion. Both Gauss-Newton and Levenberg-Marquardt methods are designed to implement NLE. Simulation model for the magnitude, phase, and noise distribution of received base band signal based on the link budgets of backscatter-radio are designed and verified with measurement states. Stack length -- the number of time samples used in the state estimation -- increases with the increase of noise level in the estimation, achieving 10-mm RMS error estimation for simulated states at different noise levels. For all variations of nonlinear state estimators, estimation accuracy using just position and velocity measurements from Received Signal Strength (RSS) and Received Signal Phase (RSP) is similar to estimation accuracy using additional measurements of acceleration, orientation from magnetometer, and angular velocity from gyroscope of the embedded IMU. The RSS-only position estimators are improved when acceleration, orientation, and angular velocity are added, suggesting that fine-scale motion capture may still be possible for non-coherent readers that cannot measure RSP. However, the coherent phase measurement appears to provide the most improvement. The absence of embedded inertial sensors would simplify the system design, allow conventional RFID tags to be used at sensor nodes, and enlarge the application range.