TL;DR
This paper tackles the challenge of unreliable GPS positioning in urban environments by combining Real-Time Kinematic (RTK) GPS with an Inertial Measurement Unit (IMU)-based pedestrian navigation algorithm on smartphones. Their approach significantly improves robustness and continuity of positioning in challenging urban canyons, demonstrating a practical solution for more reliable smartphone navigation.
What Problem Does This Paper Address?
Global Navigation Satellite Systems (GNSS), particularly Real-Time Kinematic (RTK) GPS, offer centimeter-level accuracy but struggle in urban environments. Tall buildings and structures block satellite signals, leading to frequent "GNSS outages" and unreliable positioning. This is a major limitation for applications relying on precise location tracking, especially for pedestrians navigating complex urban landscapes. While newer smartphones with dual-frequency GNSS capabilities (like the Xiaomi Mi 8) offer improvements, they still falter in these signal-scarce areas. The paper aims to bridge this gap by integrating IMU data to provide continuous and reliable positioning even when GNSS signals are lost.
Methods
The core of the paper's contribution lies in the fusion of RTK GPS with an IMU-based pedestrian navigation system. Here's a breakdown of the key techniques:
- RTK GPS Implementation: They utilize the RTKLIB software package, which employs an Extended Kalman Filter (EKF) for state estimation. This is a standard approach for processing GNSS data to achieve high accuracy. They investigate both single-frequency and dual-frequency RTK using the Xiaomi Mi 8's BCM47755 chipset.
- IMU-Based Pedestrian Navigation:
- Attitude and Heading Reference System (AHRS): They employ the Madgwick algorithm, a popular quaternion-based algorithm, to estimate the smartphone's attitude (roll, pitch, yaw) using data from the smartphone's MEMS IMU (accelerometer, gyroscope, magnetometer). This provides crucial orientation information.
- Zero Velocity Update (ZUPT): To mitigate the drift inherent in IMU-based dead reckoning, they integrate the ZUPT algorithm. This algorithm leverages the fact that a pedestrian's foot is stationary during the stance phase of walking. By detecting these zero-velocity periods, the algorithm corrects the IMU's estimated position and velocity, significantly reducing long-term drift.
- Fusion Strategy: The IMU-based pedestrian navigation algorithm (combining Madgwick for attitude and ZUPT for position/velocity correction) acts as a fallback when RTK GPS signals are lost. The system synchronizes RTK and IMU outputs to provide initial attitude for the IMU system and uses the IMU's output to fill in positioning gaps during GNSS outages.
Datasets
The paper does not explicitly mention training datasets. Instead, it focuses on collecting real-world data for evaluation.
- Data Collection: The authors collected GNSS raw measurements (pseudorange, carrier phase, etc.) and IMU measurements (accelerometer, gyroscope, magnetometer) using:
- Xiaomi Mi 8: This smartphone, equipped with the BCM47755 chipset, was used as the primary device for testing both RTK and the integrated IMU-based navigation.
- Geo++ RINEX Logger: An app used to log GNSS measurements in RINEX format.
- Sensorstream_IMU+GPS: An app used to collect IMU measurements.
- Reference Data: For comparison and validation, they used:
- ComNav Technology M300 receiver: Used as a base station for RTK.
- NovAtel OEM617-CDS-R0G-550 receiver with GPS500 antenna: Used as a reference rover for dual-frequency RTK performance comparison.
Key Insights and Ideas
- Urban Canyons are a Major Hurdle for RTK: The paper strongly reinforces that even with dual-frequency GNSS, urban environments with tall buildings severely degrade RTK performance, leading to frequent positioning losses.
- IMU Fusion is Key for Continuity: The integration of an IMU-based pedestrian navigation algorithm (specifically using Madgwick for attitude and ZUPT for drift correction) is crucial for maintaining continuous positioning during GNSS outages.
- ZUPT is Essential for Pedestrian Dead Reckoning: Without ZUPT, IMU-based dead reckoning alone would suffer from unacceptable drift over time, rendering it useless for accurate pedestrian tracking.
- Dual-Frequency Helps, But Isn't a Silver Bullet for Urban RTK Fixed Rate: While dual-frequency GNSS (L1+L5) on the Xiaomi Mi 8 improved static convergence time and reduced biases/fluctuations, it did not significantly increase the fixed rate of ambiguity resolution in dynamic urban scenarios. This suggests that antenna performance and severe multipath effects are still dominant limiting factors for achieving high fixed rates in challenging environments.
- IMU Assistance Makes RTK "More Robust and Adaptable": The core takeaway is that combining RTK with IMU-based pedestrian navigation makes the overall system significantly more reliable and capable of providing continuous positioning in complex urban environments where RTK alone would fail.
Key Results
- Static Mode: Both single-frequency and dual-frequency RTK on the Xiaomi Mi 8 could achieve fixed solutions with cm-level precision after a few minutes. Dual-frequency RTK reduced convergence time and improved the fixed rate.
- Dynamic Mode (Sports Ground):
- Dual-frequency RTK with GPS and Galileo showed the least biases and fluctuations.
- Galileo integration improved continuity by increasing the number of visible satellites.
- However, the fixed rate for dual-frequency RTK on the Xiaomi Mi 8 did not significantly improve compared to single-frequency RTK in dynamic mode, attributed to signal quality issues from the smartphone's antenna.
- Dynamic Mode (Urban Canyon):
- RTK without IMU assistance provided very few, imprecise positioning points due to GNSS outages and multipath.
- RTK with IMU assistance (ZUPT aiding INS) successfully filled in the gaps during GNSS outages, providing continuous and more reliable trajectory estimation. Figures 15 and 16 visually demonstrate this continuity.
- Madgwick Algorithm Performance: The Madgwick algorithm accurately estimated the attitude of the Xiaomi Mi 8 when tested on a turntable, showing good correspondence with the turntable's known attitude changes.
Limitations and Future Work
- Limitations:
- The fixed rate of dual-frequency RTK on the Xiaomi Mi 8 in dynamic mode did not improve significantly, likely due to the poor quality of signals from the smartphone's antenna and severe multipath effects.
- Tying smartphones to a pedestrian's foot for IMU data collection is inconvenient and not practical for everyday use.
- Future Work:
- Conduct experiments to assess accuracy and stability in a wider range of environments.
- Develop a more robust GNSS/INS integration system (loosely or tightly coupled) for smartphones to enhance anti-jamming capabilities.
- Explore alternative algorithms to assist RTK during GNSS outages, moving away from foot-mounted IMUs towards more practical handheld smartphone solutions for continuous positioning.
Summary updated