IEEE Robotics & Automation Magazine - September 2023 - 40

RADAR
In the " Object Detection " section, radar sensors were
observed to effectively recognize the objects that were not
easily identified by lidar sensors, such as glass or highly
reflective objects. Furthermore, they provided stable measurements
even in a foggy environment. Therefore, they can
be concluded as suitable for applications that operate under
harsh environments or places with a large quantity of glass or
reflective objects that affect the performance of lidar sensors.
Unlike lidar, as the radars exhibit the mechanism of detecting
objects by analyzing the wavelength, they can efficiently
detect objects such as people, cars, and bulky obstacles. In
addition, radar sensors are generally lightweight and can be
easily installed on any mobile application platform.
Although the difference between FMCW radar and pulse
radar has not been considered in this study, information on the
applicability of each radar sensor can be inferred by understanding
operation mechanisms and obtained data. The FMCW radars
such as IWR 1443 considered in this study, are often equipped
with multiple antennas for measuring the directional angles and
distances of objects. They have been widely employed in the
safety systems of autonomous vehicles. Furthermore, advanced
algorithms may be used to achieve complex detection such
as human gesture recognition. However, they produce large
amounts of data, which cause a high computational load.
As seen in X4M03, pulsed radar sensors can only measure
the distance to an object, not the directional angle, due to their
single antenna. Because the wave power of pulsed radar sensors
is relatively low, the signal-to-noise ratio is also low, making
them difficult to detect objects clearly. However, X4M03
measures the distance to an object with higher accuracy than
IWR 1443 because of its wide frequency bandwidth and small
data size. This type of radar sensor can be used for applications
such as people counting in a large supermarket and traffic
control to measure the speed of moving cars.
OVERALL COMPARISON AND
COMPLEMENTARY STRATEGIES
Each lidar and radar sensor has its own set of advantages and disadvantages
that differ among sensors. It is therefore essential that
sensor characteristics are properly investigated before choosing
sensors suited to specific applications. The experimental results in
the previous section indicate several evaluation indices. In summary,
the advantages, disadvantages, suitable applications, and
quantitative comparison of each sensor in terms of six criteria are
presented in Table 4. The six criteria include measurement precision,
distance range, environmental robustness, sensor size, cost,
and data dimension. For each criterion, points were uniformly
assigned between 0 and 5.
By considering the advantages and disadvantages of each
sensor, it would be beneficial to appropriately combine multiple
sensors to ensure that they complement each other. In particular,
when recognizing objects, lidar sensors obtain high-resolution
point cloud data that reflect their accurate positions, unlike radar
sensors; however, they suffer from degraded detection ability in
the case of certain materials, such as glass. This limitation can
40 IEEE ROBOTICS & AUTOMATION MAGAZINE SEPTEMBER 2023
be overcome by using radar sensors. Furthermore, the material
types or the speed of dynamic objects can also be estimated
using radar sensors. As another sensor fusion application, a
robust outdoor autonomous driving system can be implemented
using a combination of lidar and radar sensors to provide robust
performance, even in a foggy environment.
In Figure 9, the operational details of four real-life applications
are presented to demonstrate the effectiveness of lidar and
radar sensors toward their combined functionality in a complementary
fashion by using the advantages and limitations of each
sensor, as discussed previously. These example applications
would provide an effective guide toward the use and combination
of sensors for achieving the desired purposes. As the first
example, lidar maps can be combined with dynamic objects recognized
by radar sensors [see Figure 9(a)]. Such an augmented
combination provides the benefit of simultaneously representing
the static point cloud from lidar sensors and dynamic information
of objects from radar sensors in the same coordinate system.
One of the benefits is that it is highly effective in avoiding
collisions with moving or fixed transparent objects. Similarly,
the second example in Figure 9(b) shows that the acrylic box is
detected by a radar sensor and displayed on the lidar map. It is
noteworthy that the acrylic box was not detected by a lidar sensor.
Next, radar sensors can detect objects behind walls, unlike
lidar sensors, as shown in Figure 9(c). The data points from
radar sensors are useful for obtaining additional features when
mapping and localizing in combination with lidar sensors, especially
in symmetric and simple geometric environments where
the latter do not function well. However, due to lower resolution
of the radar sensor, a combination of the two sensors is recommended
for a synergistic effect. Finally, multiple roles in a
mobile robot system can be assigned to different lidar sensors.
The autonomous mobile robot shown in Figure 9(d) is designed
for surveillance and defect detection in tunnels. It has Velodyne
VLP-16 and RPLIDAR-A2 lidar sensors that fulfill the functions
of global localization and avoidance of the lateral drainage
path by scanning the field directly ahead and also at an angle
of 45° below, respectively. By extracting the clustered line from
the scan data of an RPLIDAR-A2 lidar sensor, the current heading
of a mobile robot relative to the aisle can be calculated to
control the driving direction. Simultaneously, a 3D mechanical
Velodyne VLP-16 lidar sensor installed on the top of the robot
scans the environment to calculate the current position according
to the global map.
CONCLUSION
With a focus on exploring synergistic combinations of commercial
sensors, this article presented a quantitative capability comparison
for representative lidar and radar sensor technologies in
terms of object detection, mapping, and environmental robustness.
Experimental test beds were specially constructed to
ensure unbiased and objective comparisons in real environments.
Considering all experimental results together, we presented
visually informative metrics that are to be considered when
selecting lidar and radar sensors for robotic applications on a
comprehensive basis. The proposed metrics will provide a useful
IEEE Robotics & Automation Magazine - September 2023

Table of Contents for the Digital Edition of IEEE Robotics & Automation Magazine - September 2023
