Page 23 - 卫星导航2021年第1-2合期

El‑Sheimy and Li Satell Navig (2021) 2:7 Page 13 of 23

Dead-reckoning techniques by fnding periodical characteristics in accelerometer and
Te basic principle of DR technology is to derive the cur- gyro measurements (Alvarez et al. 2006), while the step
rent navigation state by using the previous navigation length is commonly estimated by training a model that
state and the angular and linear movements. Te angu- contains walking-related parameters (e.g., leg length and
lar and linear movements can be obtained by using the walking frequency) (Shin et al. 2007).
measurements of sensors such as inertial sensors, cam- Tere are DR algorithms based on other types of sen-
eras, magnetometers, and odometers. Among them, sors, such as visual odometry (Scaramuzza and Fraun-
inertial sensors are most widely used for DR. Tere are dorfer 2011) and wheel odometry (Brunker et al. 2018).
two main DR algorithms based on inertial sensors: INS Magnetometers (Gebre-Egziabher et al. 2006) are also
mechanization and PDR. Te former is widely used in used for heading determination.
land-vehicle, airborne, and shipborne PLAN applica- To achieve a robust long-term DR solution, there are
tions, while the latter is a common method for pedestrian several challenges, including the existence of sensor
navigation. Figure 6 shows the fow of the INS mechani- errors (Li et al. 2015), the existence of the misalignment
zation and PDR algorithms. INS can provide 3D naviga- angle between device and platform (Pei et al. 2018), and
tion results, while PDR is a 2D navigation method. the requirement for position and heading initialization.
Te INS mechanization works on the integration of Also, the continuity of data is very important for DR. In
3D angular rates and linear accelerations (Titterton et al. some applications, it is necessary to interpolate, smooth,
2004). Te gyro-measured angular rates are used to or reconstruct the data (Kim et al. 2016).
continuously track the 3D attitude between the sensor DR has become a core technique for continuous and
frame and the navigation frame. Te obtained attitude seamless indoor/outdoor PLAN due to its self-contained
is then utilized to transform the accelerometer-meas- characteristics and robust short-term solutions. It is
ured specifc forces to the navigation frame. Afterward, strong in either complementing other PLAN techniques
the gravity vector is added to the specifc force to obtain when they are available or bridging their signal outages
the acceleration of the device in the navigation frame. and performance-degradation periods.
Finally, the acceleration is integrated once and twice
to determine the 3D velocity and position, respectively. Database-matching techniques
Terefore, the residual gyro and accelerometer biases in Te principle for database matching is to compute the
general cause position errors proportional to time cubed diference between the measured fngerprints and the
and time squared, respectively. reference fngerprints in the database and fnd the closest
In contrast, the PDR algorithm (Li et al. 2017) deter- match (Li et al. 2020a). Database-matching techniques
mines the current 2D position by using the previous are used to process data from various sensors, such as
position and the latest heading and step length. Tus, it cameras, LiDAR, wireless sensors, and magnetometers.
consists of platform-heading estimation, step detection, Te database-matching process consists of the steps of
and step-length estimation. Te platform heading is usu- feature extraction, database learning, and prediction.
ally calculated by adding the device-platform misalign- Figure 7 demonstrates the processes. First, valuable fea-
ment (Pei et al. 2018) into the device heading, which can tures are extracted from raw sensor signals. Afterward,
be tracked by an Attitude and Heading Reference System features at multiple reference points are combined to
(AHRS) algorithm (Li et al. 2015). Te steps are detected

Fig. 7 Diagram of database matching process
Fig. 6 Diagram of INS mechanization and PDR algorithms