Innovative technologies will change our life upside down. 3D/HD mapping is a fundamental aspect for autonomous driving.
Naver Labs and 3D Mapping Project
Naver
Labs is an ambient intelligence company located in Gyeonggi that seeks to develop technology that
is able to provide the information and services users require before it is
requested. They also provide solutions that can be seamlessly integrated into
everyday lives through location technology and intellectualizing the mobility
that connects different spaces.
They
have multiple projects including 5G Brainless Robots, indoor 3D/HD mapping
robots, machine-readable HD maps, indoor AR navigation and many more that are constantly being updated and transformed to create a better and more
technologically-advanced world.
Naver Labs x Hyundai Department Store (Pangyo)
It is very easy to get
lost in large indoor spaces such as shopping malls or department stores. The
vast number of stores as well as the large crowds will leave you clueless as to
where you are even if you are looking for stores that you have been to before.
Instead of looking for the information desk or a map, directions via augmented
reality (AR) navigation can be extremely useful in situations like these.
 |
| M-1M1 is an indoor 3D/HD mapping robot that navigates autonomously in indoor spaces ⓒNaver Labs Corp. |
In order to effectively
utilize AR, the location of the user first needs to be determined. There are
many obstacles to determining a user’s location with accuracy; many people
crowded in the same location, strong contrasting lights and rapidly changing
event structures form a large part of the hurdles faced.
Naver Labs has developed a
new technology that they have showcased at the multi-floor shopping center of
Hyundai Department Store in Pangyo, Gyeonggi-do, South Korea. The indoor AR
navigation demo showcases an accurate and stable navigation system in a real
environment, which is the goal that they have challenged themselves to achieve
by overcoming technological limitations.
A few factors were
utilized in this technology:
1.
Integration of Complex
Spatial Data
Visual localization (VL),
which plays an important role in indoor AR navigation, begins with spatial
information collected by a mapping robot. To provide seamless services for
stairs, multiple floors and narrow hallways in a real environment, it is
essential to have spatial information that accurately integrates information
from complex spaces inaccessible to a mapping robot. To circumvent this,
backpack-like mapping equipment developed in-house was used to successfully
link vertical data in the outdoor and multi-floor spaces of the mall.
2.
Real-Time Location
Correction
The VL technology of Naver
Labs accurately identifies current locations even in indoor spaces where GPS
systems cannot work with a high degree of precision. In situations where the
user has to keep moving, however, it is difficult to augment the content in a
stable manner due to slight location deviations occurring in real time. To
overcome this problem, Naver Labs applied visual-inertial odometry (VIO), which
tracks locations in real time by analysing sensor and image information.
Nevertheless, content
often lags by several meters compared to real time locations due to rapid
camera movements, network delays and temporary deviations. Naver Labs has
solved many of these problems with a real-time camera developed in-house that
aids tracking. It uses a location-correction algorithm for precise real-time
positioning, which can augment and track content in a stable and accurate
manner even while the user is moving by combining various filtering logic
approaches and a location-prediction technique that analyses previous traveling
routes and calculates current locations.
3.
Seamless AR
Navigation Between Floors
Achieving seamless AR
navigation between floors was a very challenging project that required Naver
Labs to identify not only spatial data for all floors, but also the precise
location of the user on the floor and to link that information with the next
scenario. This is simple for humans, but not for machines.
In order to connect
directions flawlessly across multiple floors, including getting on and off
escalators, Naver Labs needed various location-based technologies. The AR
navigation system identifies current locations through visual localization
(VL), tracks the location in real-time and determines the user’s surrounding
environmental elements via camera and visual object tracking (VOT), and
thereafter augments content to determine an accurate location.
4.
AR and Real Space
Connection via 3D Maps
Finally, Naver Labs was
able to develop a location-based AR authoring system mediated by a 3D map made
by a mapping robot. As the system has the same coordinates as real spaces, the
content appears at an accurate location if you set the content onto the 3D map.
Not only can it pinpoint the correct location values accurately, but specific
content parameters and interactions can easily be applied to these real spaces.
 |
| Actual pictures of the real space ⓒNaver Labs Corp. |
 |
| Turns into a photo-realistic indoor 3D map ⓒNaver Labs Corp. |
You are even able to
browse the 3D map and augment locations you want without even leaving your
home. This opens up the doors to a wide variety of uses. This demo also applies
to various possible scenarios such as interactively opening a new menu in a
store or interactive promotional information.
MobilTech - The Leader of Lidar Technology
Another company that has
been developing a 3D mapping service is MobilTech, a 2019 Korea Mobility
Innovation start-up. This start-up, which operates in autonomous driving 3D
mapping and cognitive fields, is recognized for its potential in those fields
and in the industry. MobilTech was founded in April 2017 and, in the relatively
short time of two and a half years, was chosen by multiple large companies with
reputations for finding promising start-ups. Companies such as Hyundai Motor,
Naver, and Woori Bank have invested in MobilTech’s development potential.
 |
| MobilTech received a lot of attention in CES 2020 |
Before heading into the
autonomous driving 3D mapping field, CEO Kim Jae-seung had studied drones. He
predicted that drones, his main research subject at that time, would occupy a
large portion of the future of the mobility market. He was convinced that the
development of 3D maps to support the autonomous driving of drones would also
be in the spotlight.
In order to develop
various services using drones in urban areas, it was determined that a 3D map
and a route planning method including a height concept derived from an existing
planar map were needed. Upon realizing this, Kim decided to start MobilTech.
After conducting more market research, he ultimately concluded that the
autonomous drone market still needs more time before it can fully be
integrated.
Instead, there was
confidence that the era of autonomous vehicles would arrive ahead of drones.
From that point, Kim turned his eyes to autonomous vehicles instead of drones
and began to delve into related fields. In particular, there were not enough
high-precision spatial information companies in the autonomous driving industry
in Korea. Since competition was less intense, he was confident of succeeding
should he enter the market first. After determining his business direction, Kim
introduced MobilTech to the world.
After the company was
founded, it undertook research and development on how to quickly and accurately
collect high-precision spatial information. It focused on collecting map data
in downtown areas through a fusion of Lidar, cameras and satellite navigation
systems (GPS). The development process had a lot of ups and downs and, just
like any other technology start-up, the large sum of money going into research
and development was a huge obstacle.
Currently, MobilTech has a
technology specialized in 3D precision mapping using a Lidar sensor that is
recognized by the industry. Of particular note, the 3D scanning “Replica
Series” developed with their own technology is one of their leading
achievements. The replica series is an integrated solution that utilizes Lidar
and cameras to collect 3D precise map data and process the data collected
through deep learning.
Perfect autonomous driving
can only be realized by combining several technologies. MobilTech is preparing
for the future with the goal of becoming one of the best in the fields of 3D
map production and autonomous driving recognition.
COMMENTS