Page 20 - 卫星导航2021年第1-2合期

El‑Sheimy and Li Satell Navig (2021) 2:7 Page 10 of 23

Global navigation satellite system (as an initializer) Techniques and algorithms for indoor navigation
GNSS localizes a receiver using satellite multilateration. Te PLAN techniques include position-fxing, Dead-
It is one of the most widely used and most well-com- Reckoning (DR), database matching, multi-sensor fusion,
mercialized PLAN technology. Standalone GNSS and and motion constraints. Figure 4 demonstrates the
GNSS/INS integration are the mainstream PLAN solu- indoor PLAN techniques. Te details are provided in the
tions for outdoor applications. In autonomous driving, following subsections.
the GNSS transfers from the primary PLAN sensor to
the second core. Te main reason is that GNSS signals Position-fxing techniques
may be degraded in urban and indoor areas. Even so, Geometrical position-fxing methods have been widely
high-precision GNSS is still important to provide an applied over the past few decades, especially in the feld
initial localization to reduce the searching space and of satellite positioning and wireless sensor networks. Te
computational load of other sensors (e.g., HD map and basic principle is the geometric calculation of distance
LiDAR) (Levinson et al. 2007). and angle measurements. By the type of measurement,
Te previous boundaries between high-precision pro- position-fxing methods include range-based (e.g., multi-
fessional and mass-market GNSS uses are blurring. A lateration, min–max, centroid, proximity, and hyperbolic
piece of evidence is the integration between high-pre- positioning), angle-based (e.g., multiangulation), and
cision GNSS techniques and mass-market chips. Also, angle-and-range-based (e.g., multiangulateration). Fig-
the latest smartphones are being able to provide high- ure 5 shows the basic principle of these methods.
precision GNSS measurements and PLAN solutions.
Table 8 lists the main GNSS positioning techniques. Range‑based methods
Single Point Positioning (SPP) and Diferential-GNSS Te location of a device can be estimated by measuring
(DGNSS) are based on pseudo-range measurements, its distance to at least three base stations (or satellites)
while Real-Time Kinematic (RTK), Precise Point Posi- whose locations are known. Te most typical method is
tioning (PPP), and PPP with Ambiguity Resolution multilateration (Guvenc and Chong 2009), which is geo-
(PPP-AR) are based on carrier-phase measurements. metrically the intersection of multiple spheres (for 3D
DGNSS and RTK are relative positioning methods positioning) or circles (for 2D positioning). Also, the
that mitigate some errors by diferencing measure- method has several simplifed versions. For example, the
ments across the rover and base receivers. In contrast, min–max method (Will et al. 2012) computes the inter-
PPP and PPP-AR provide precise positioning at a sin- section of multiple cubes or squares, while the centroid
gle receiver by using precise satellite orbit correction, method (Pivato et al. 2011) calculates the weighted aver-
clock correction, and parameter-estimation models. age of multiple base station locations. Moreover, the
Tey commonly need minutes for convergence (Trim- proximity method (Bshara et al. 2011) is a further simpli-
ble 2020). fcation by using the location of the closest base station.
Tere are other types of PLAN sensors, such as mag- Meanwhile, the diferences of device-base-station ranges
netometer, odometer, UWB, ultrasonic, and pseudolite. can be used to mitigate the infuence of device diversity
In recent years, there appears relatively low-cost UWB and some signal-propagation errors (Kaune et al. 2011).
and ultrasonic sensors (e.g., (Decawave 2020; Marvel- For position-fxing, the base station location is usually
mind 2020). Such sensors typically can provide a deci- set manually or estimated using base-station localization
meter-level ranging accuracy within a distance of 30 m. approaches (Cheng et al. 2005). Te distances between
Also, Apple has built a UWB module into the iPhone 11, the device and the base stations are modeled as Path-Loss
which may bring new opportunities for indoor PLAN. To Models (PLMs) and parameters are estimated (Li 2006).
summarize, Table 9 illustrates the principle, advantages, To achieve accurate ranging, it is important to mitigate
and disadvantages of the existing PLAN sensors. the infuence of error sources (e.g., ionospheric errors,

Table 8 GNSS positioning techniques
Technique Accuracy Measurement Methods for accuracy improvement

SPP Meter‑level Pseudo‑range None
DGNSS Decimeter‑level to meter‑level Pseudo‑range Diferential measurements
RTK Centimeter‑level Carrier‑phase Diferential measurements, ambiguity resolution
PPP Decimeter‑level to centimeter‑level Carrier‑phase Precise satellite orbit and clock corrections, foat ambiguity
PPP‑AR Similar to RTK Carrier‑phase Precise satellite orbit and clock corrections, ambiguity resolution