believe when more road test data is accumulated and analyzed, the accuracy
and efficiency of these vehicles will be
How Meituan Uses the Vehicles
in Delivery Scenarios
Meituan is the world’s largest e-commerce platform for local services.
Meituan’s service covers over 200 categories, including catering, on-demand
delivery, car sharing, bicycle sharing,
hotel and travel, movie, entertainment
and lifestyle, and spreads over 2,800
counties, districts, and cities in China.
In 2017, Meituan served 310 million active consumers and 4. 37 million active
merchants on the platform.
One of Meituan’s services is its
on-demand food delivery known as
Meituan Waimai. By May 2018, Meituan Waimai was delivering 21 million
orders per day and had hired 600,000
food-delivery riders. The service is usually within three kilometers, with tight
time limits of 30 minutes. The fulfillment process includes three phases:
1. The courier goes to the restaurant
to pick up the food, usually by walking
through a shopping mall to get to the
restaurant; 2. The courier transports
the food next to consumer’s building;
3. The courier walks or takes an elevator
to the consumer (See Figure 6). In practice, each phase takes approximately
one-third of the total delivery time.
Different phases need different
types of vehicles. In phase 1 and phase
3, we use an indoor robot, as shown in
Figure 7. This robot is 0.5m by 0.8m; its
small size allows for easy entrance to
shopping malls and office buildings. It
does localization based on WiFi fingerprint and vision SLAM. It can also communicate through Zigbee to an elevator control module, thus can go up and
down the buildings. The robot receives
order information from the cloud
scheduling system, runs to the merchant following the scheduled route,
opens the top cover automatically so
the merchant can put the food inside.
When approaching its destination, the
robot sends a text message to the user’s
mobile app, and then the user can pick
up the food using the password code
included in the text message.
In phase 2, a larger and faster autonomous delivery vehicle is used for street
transportation, as shown in Figure 8.
The vehicle measures one-meter wide
and two-meters long, with maximum
speed of 40km/h and maximum load of
10 orders. It uses the same technology
as an autonomous passenger vehicle,
including lidar, camera, and GNSS receiver. It can detect pedestrians, bicycles, automobiles, and other obstacles,
and can also react to traffic lights.
Challenges and Opportunities
There are many challenges for the
large-scale deployment of autonomous
delivery vehicles: The technology is not
mature yet, the entire ecosystem must
be further developed to make it more
reliable, and the costs much shrink.
Moreover, our living infrastructure
is not yet ready for autonomous driving. Many communities have locked or
gated entrances, which require manual
operation using a key or access card.
Many buildings have revolving or swing
doors, which are easy for humans to use
but very difficult for robots. Elevators
are rarely robot-ready. In fact, we must
talk to building owners to get a permit
to install a communication module in
every elevator. These frustrations must
be handled before we can fully enjoy
autonomous vehicle deployment.
Government regulation of autono-
mous vehicle is a critical concern.
While these delivery vehicles run at
a fairly slow speed, most regulators
consider “slow” a grey area, fitting be-
tween high-speed passenger vehicles
and bicycles. Government regulations
are not ready to handle pure level 4
(driver-free) vehicles like autonomous
delivery vehicles. What happens if the
autonomous vehicle is involved in an
accident or a traffic violation? Who
is responsible? Closed environments
such as common in last-mile deliv-
ery can be used as pilot scenarios for
learning that enable more complex,
open road scenarios. This suggests
two autonomous driving vehicles de-
velopment methodologies: find a kill-
er-level use case to drive the business
model; or find the best technology and
build the ecosystem.
Fortunately, this somewhat immature technology is acceptable for slow-speed delivery vehicles. The lack of
infrastructure support may prevent us
from mass deployment in some situations, but there are still many suitable
scenarios for first-stage deployment.
As for government regulation, the Chinese government is among the most
supportive for high-tech innovations
like autonomous vehicles.
The Future of Autonomous
China’s e-commerce boom brings a
huge volume of logistics demand, both
for express package delivery and for
on-demand food delivery. The last-mile
delivery is a perfect use case for autonomous driving technology.
Large-scale deployment of autonomous vehicles still depends on technology maturity and governing regulations.
Nevertheless, these issues are not showstoppers. There are many pilot scenarios with government support, helping
the industry step into the water of autonomous driving in order to accumulate the data and real-world experience
needed to improve its technology.
1. Levine, S. What it really costs to turn a car into a self-driving vehicle; https://qz.com/924212, 2017.
2. Xinhua News. Chinese express firms deliver over 40
bln parcels in 2017, 2018; https://bit.ly/2QnioCi
3. Xinhua News. China’s express delivery sector prepares
for post-holiday bonanza, 2018; http://www.xinhuanet.
Huaxia Xia is Scientist and General Manager of the
Autonomous Delivery Department at Meituan, Beijing.
Haiming Yang is Chief Architect at JD CTO Group, Beijing.
Copyright held by owners/authors.
Publication rights licensed to ACM. $15.00..
Figure 7. Meituan’s indoor delivery robot.
Figure 8. Meituan’s outdoor delivery