Page 25 - 卫星导航2021年第1-2合期

El‑Sheimy and Li Satell Navig (2021) 2:7 Page 15 of 23

Table 11 PLAN sensors corresponding to diferent senses extra hardware costs. For land-based vehicles, the Non-
of the human body Holonomic Constraints (NHC) can improve the head-
ing and position accuracy signifcantly when the vehicle
Human sensing Smartphone Vehicle
moves with enough speed (Niu et al. 2010), while the
Eyes Camera Camera, LiDAR Zero velocity UPdaTe (ZUPT) and Zero Angular Rate
Ears GNSS, WiFi, BLE, UWB GNSS, RADAR, ultrasonic, Update (ZARU, also known as Zero Integrate Head-
(new), 5G (future), UWB, 5G (future), ing Rate (ZIHR)) respectively provide zero-velocity and
microphone microphone
Moving direction Gyro, magnetometer Gyro, magnetometer zero-angular-rate constraints when the vehicle is quasi-
static (Shin 2005). When the vehicle moves at low speed,
Moving distance Accelerometer Odometer, accelerometer a steering constraint can be applied (Niu et al. 2010).
Memory Map HD map, map Moreover, there are other constraints such as the height
Decision making Processor, algorithm Processor, algorithm
constraint (Godha and Cannon 2007) and the four-wheel
constraint (Brunker et al. 2018).
For pedestrian navigation, ZUPT (Foxlin 2005) and
To be specifc, for position-fxing and database-match-
ing methods, the loss of signals or features lead to out- ZARU (Li et al. 2015) are most commonly used. Also, the
NHC and step velocity constraint (Zhuang et al. 2015)
ages in the PLAN solution. Also, changes in the model have been applied. Furthermore, in indoor environments,
and database parameters may degrade the PLAN per- constraints such as the corridor-direction constraint
formance. To mitigate these issues, DR techniques can (Abdulrahim et al. 2010), the height constraint (Abdulra-
be used (El-Sheimy and Niu 2007a, b). Moreover, the use him et al. 2012), and the human-activity constraint (Zhou
of other techniques can enhance position-fxing through et al. 2015) are useful to enhance the PLAN solution.
more advanced base station position estimation (Cheng
et al. 2005), propagation-model estimation (Seco and
Jiménez 2017), and device diversity calibration (He et al. Use cases
2018). Also, the number of base stations required can be Multi-sensor-based indoor navigation has been utilized
reduced (Li et al. 2020b). On the other hand, position- in various applications, such as pedestrians, vehicles,
fxing and database-matching techniques can provide robots, animals, and sports. Tis chapter introduces
initialization and periodical updates for DR (Shin 2005), some examples. Tree of our previous cases on indoor
which in turn calibrate sensors and suppress the drift of navigation are demonstrated. Te used vehicle platforms
DR results. include smartphones, drones, and robots.
Database matching can also be enhanced by other tech-
niques. For example, the position-fxing method can be Smartphones
used to reduce the searching space of database-match- Tis case uses an enhanced information-fusion struc-
ing (Zhang et al. 2017b), predict the database in unvis- ture to improve smartphone navigation (Li et al. 2017).
ited areas (Li et al. 2019d), and predict the uncertainty Te experiment uses the built-in inertial sensors, WiFi,
of database-matching results (Li et al. 2019e). Also, a and magnetometers of smartphones. By combining the
more robust PLAN solution may be achieved by integrat- advantages of PDR, WiFi database matching, and mag-
ing position-fxing and database-matching techniques netic matching, a multi-level quality-control mechanism
(Kodippili and Dias 2010). is introduced. Some quality controls are presented based
From the perspective of integration mode, there are on the interaction of sensors. For example, wireless posi-
three levels of integration. Te frst level is loosely cou- tioning results are used to limit the search scope for mag-
pling (Shin 2005), which fuses PLAN solutions from dif- netic matching, to reduce both computational load and
ferent sensors. Te second level is tightly-coupling (Gao mismatch rate.
et al. 2020), which fuses various sensor measurements to Te user carried a mobile phone and navigated in a
obtain a PLAN solution. Te third level is ultra-tightly- modern ofce building (120 m by 60 m) for nearly an
coupling, which using the data or results from some sen- hour. Te smartphone has experienced multiple motion
sors to enhance the performance of other sensors. modes, including handheld horizontally, dangling with
hand, making a call, and in a trouser pocket.
Te position results are demonstrated in Fig. 8. When
Motion constraints directly fusing the data from PDR, WiFi, and magnetic
Motion constraints are used to enhance PLAN solu- in a Kalman flter, the results sufer from large posi-
tions from the perspective of algorithms, instead of add- tion errors. Te ratio of large position errors (greater
ing extra sensors. Such constraints are especially useful than 15 m) reached 33.4%. Such a solution is not reli-
for low-cost PLAN systems that are not afordable for able enough for user navigation. By using the improved