![](https://csdnimg.cn/release/download_crawler_static/12225820/bg4.jpg)
new setup method for LiDAR is firstly introduced. Und er
this new setup, the density of LiDAR data has been greatly
improved, and the distance between two adjacent scan
points would be much s horter. Hence, the influence of the
road’s bumpiness would be greatly cut down. Then, by
analyzing the scan line of LiDAR, a mathematical model
is deduced to simulate the distribution of these real scan
points on the same scan line. An FMF-based obstacle
detection algorithm is presented by fusing multi-cues and
matching with simulated scan lines. The experimental
results show that the proposed algorithm is stable, effec-
tive, and robust.
A new setup method of compact LiDAR
Recently, 3-D LiDARs have become more and more
important for detecting obstacles by ALVs. In current
mainstream ALV teams, including Stanford’s Junior,
28
KIT’s AnnieWAY,
26
Google self-driving car, the vehicle
used in the study of Ha¨selich et al.,
29
and our team
3
(shown in Figure 2), the velodyne HDL-64E LiDAR is
widely employed for obstacle detection. In these applica-
tions, their 3-D LiDARs are all uprightly fixed on their
ALV platforms.
The shortcoming of uprightly fixed LiDAR
In the aforementioned applications, 3-D LiDAR is always
fixed uprightly on its ALV platform. There is a serious
shortcoming: In this uprightly setup, LiDAR cannot view
the region around the vehicle, which is sheltered from the
body of vehicle itself. The vehicle would not be suffered
from the blind area when it drove in flat environments or
structured environments, since roads in these scenarios are
structured and flat. The borders of the road on two sides
could also help to keep the vehicle safe.
30
However, in a
field environment, roads are usually narrow and unstruc-
tured with stakes and pits. Thus, the blind area would cause
serious damages to the vehicle.
Under the traditional upright setup, the blind region
around the vehicle can be analyzed as follows (Figure 3):
The distance from the blind region to the side of the vehicle
is B
1
¼ W
1
(H
1
þ H
2
)/H
1
W
1
, where W
1
is the half-
width o f the vehic le, H
1
is the height of the mounted
LiDAR, and H
2
is the height of the vehicle. The size of the
blind region in front of the vehicle is B
2
¼ W
2
(H
1
þ H
2
)/
H
1
W
2
, where W
2
is the distance from the LiDAR to the
head of the vehicle. The blind region is shown in Figure 3.
The details of the proposed new setup
Considering the shortcoming brought by traditional
upright setup, in order to make a full use of LiDAR sensor,
this article presents a n ew setup, that is, to make the
LiDAR tilt on the s ide of the vehicle. The introduced new
setup is described in Figure 1: Two compacted 3-D
LiDARs are fixed on each side of the vehicle top—one
is toward the left and another is toward the right. Both are
equally angled forward.
There are three rules to obey for this new setup: (1)
Make sure that no lasers from two LiDARs illuminate each
other directly; otherwise, the sensors will be damaged; (2)
the visual range around the vehicle depends on (Figure
4(a)); and (3) the visual range in front of the vehicle and the
region of overlapped scans depends on (Figure 4(b)). The
visual range with the proposed setup is shown in Figure 4,
in which angles and their corresponding visual ranges are
shown on two viewpoints: a front view and a top view. The
Figure 2. 3-D LiDARs are all uprightly fixed on current mainstream ALVs. ALVs: autonomous land vehicles.
W
2
H
1
(a) (b)
W
1
H
2
B
1
B
2
Figure 3. Analysis on the blind area between two setups of 3-D LiDAR: (a) side view and (b) front view.
4 International Journal of Advanced Robotic Systems