Page 13 - 卫星导航2021年第1-2合期

El‑Sheimy and Li Satell Navig (2021) 2:7 Page 3 of 23

Te research work (Basnayake et al. 2010) shows the Te cost of indoor PLAN applications depends on the
accuracy requirements in Vehicle-to-Everything (V2X) sensors used. Te main sensors and solutions will be
applications for which-road (within 5 m), which-lane introduced in the following section.
(within 1.5 m), and where-in-lane (within 1.0 m). Te
National Highway Safety Administration (NHTSA 2017) Main players of indoor navigation
reports a requirement of 1.5 m (1 sigma, 68% probabil- Various researchers and manufacturers investigate
ity) tentatively for lane-level information for safety appli- indoor PLAN problems from diferent perspectives.
cations. Te research work (Reid et al. 2019) derives an Table 3 lists the selected research works that can refect
accuracy requirement based on road geometry standards the typical navigation accuracy for diferent sensors,
and vehicle dimensions. For passenger vehicle operat- while Table 4 shows the selected players from the indus-
ing, the bounds of lateral and longitudinal position errors trial. Te primary sensor, reported accuracy, and sensor
are respectively 0.57 m (95% probability in 0.20 m) and costs are covered.
1.40 m (95% probability in 0.48 m) on freeway roads, and Te actual PLAN performance is related to the factors
both 0.29 m (95% probability in 0.10 m) on local streets. such as infrastructure deployment (e.g., sensor type and
In contrast, the research work (Levinson and Trun deployment density), sensor grade, environment factors
2010) believes that centimeter positioning accuracy (with (e.g., the signifcance of features and area size), and vehi-
a Root Mean Square (RMS) error of within 10 cm) is suf- cle dynamics.
fcient for public roads, while the report (Agency 2019) In general, diferent types of sensors have various prin-
defnes the accuracy for autonomous driving to be within ciples, measurement types, PLAN algorithms, perfor-
20 cm in horizontal and within 2 m in height. Meanwhile, mances, and costs. It is important to select the proper
the research work (Stephenson 2016) reports that active sensor and PLAN solution according to requirements.
vehicle control in ADAS and autonomous driving appli-
cations require an accuracy better than 0.1 m. Beyond
research, the goal for autonomous driving is set at the State of the art
centimeter-level by many autonomous-driving compa- To achieve an accurate and robust PLAN for autonomous
nies (e.g., (Nvidia 2020)). To summarize, autonomous vehicles, multiple types of sensors and techniques are
driving requires the PLAN accuracy at decimeter-level required. Figure 1 shows part of the PLAN sensors that
to centimeter-level. Te current cost is in the order of $ have been in autonomous cars. Tis section summarizes
1000 to $ 10,000 (when using three-Dimensional (3D) the state-of-the-art sensors and PLAN techniques.
Light Detection and Ranging (LiDAR)).
For indoor mapping, the review paper (Cadena et al. Sensors for indoor navigation
2016) shows that the accuracy within 10 cm is sufcient Te sensors include environmental monitoring and
for two-Dimensional (2D) Simultaneous Localization and awareness sensors (e.g., HD map, LiDAR, RAdio Detec-
Mapping (SLAM). Indoor mapping is commonly con- tion and Ranging (RADAR), camera, WiFi/BLE, 5G,
ducted with a vehicle that moves slower in a smaller area and Low-Power Wide-Area Network (LPWAN)), and
when compared with autonomous driving. Te cost of a the navigation sensors (e.g., Inertial Navigation Systems
short-range 2D LiDAR for indoor mapping is in the order (INS) and GNSS). Te advantages and challenges for
of $ 1000. each sensor are also introduced and compared.
Te research work (Rantakokko et al. 2010) illustrates
that frst responders require indoor PLAN accuracy of Environmental monitoring and awareness sensors (aiding
1 m in horizontal and within 2 m in height. Te cost for sensors for navigation system)
frst responders is at the $ 1,000-level. HD maps
For mass-market applications, it is difcult to fnd a Car-mounted road maps have been successfully com-
standard of PLAN accuracy requirement. An accepted mercialized since the beginning of this century. Also,
accuracy classifcation is that 1–5 m is high, 6–10 m is companies such as Google and HERE have launched
moderate, and over 11 m is low (Dodge 2013). Te verti- indoor maps for public places. Tese maps contain roads,
cal accuracy requirement is commonly on the foor-level. buildings, and Point-of-Interest (POI) information and
For such applications, it is important to use existing con- commonly have meter-level to decimeter-level accuracy.
sumer equipment and reduce base station deployment Te main purpose of these maps is to assist people to
2
costs. On average, the deployment in a 100 m -level area navigate and perform location service applications. Te
costs approximately $ 10-level. Te E-911 cellular emer- main approaches for generating these maps are satel-
gency system uses cellular signals and has an accuracy lite imagery, land-based mobile mapping, and onboard
requirement of 80% for an error of 50 m (FCC 2015). GNSS crowdsourcing.