Alibaba Logistics Vehicle Xiaomanlv Overturned

Author: Neuvition, IncRelease time:2022-03-29 01:55:57

In the field of last-mile distribution, Alibaba launched the logistics vehicle Xiaomanlv for express delivery in the “last-three kilometers”, and put 350 Xiaomanlv logistics vehicles into use in more than 200 colleges and universities in China, successfully moving towards large-scale commercial use.

Logistics Vehicle Xiaomanlv Overturned

Two weeks ago, logistics vehicle Xiaomanlv had an accident in a college. The logistics vehicle Xiaomanlv drove directly into a piece of wet cement that had been filled recently. Xiaomanlv was firmly stuck in it, unable to move and work normally. It is reported that this is not the first time that Xiaomanlv has encountered a predicament. For example, when the road is blocked by bicycles or road repairs fence, Xiaomanlv can only wait in place to ask for rescue. Another failure situation happens when there are too many people around, Xiaomanlv will be trapped and unable to move an inch.

The long tail scenario problem of logistics vehicle autonomous driving

Logistics vehicle Xiaomanlv was designed and developed by the Autonomous Driving Laboratory of Alibaba Damo Academy. In terms of hardware, the Xiaomanlv is equipped with a customized 32-line LiDAR at the front and rear. There are six cameras around the car to form a surround-view solution, as well as sensors such as millimeter-wave radar and inertial navigation. In terms of software, Xiaomanlv integrates artificial intelligence and autonomous driving technology, with human-like cognitive intelligence, and the brain’s emergency response speed is seven times that of humans. It only takes 0.01 seconds to discriminate the action intentions of more than 100 pedestrians and vehicles. In addition, it has a five-fold redundant safety system to ensure that automated logistics vehicle runs safely. 

Logistics vehicle Xiaomanlv can run 100 kilometers on a charge of four degrees of electricity and can deliver up to 500 couriers per day. And it can work under rough weather: thunderstorm lightning, high temperature, rain, and snow.

With the help of multiple sensors, logistics vehicle Xiaomanlv can actively avoid obstacles. But it still turned over in this wet cement field.

The analysis says that although the automated logistics vehicle Xiaomanlu is driving in a relatively closed environment, most driving routes and stops are pre-planned. But the wet cement land has not been updated to the map library of Xiaomanlv, so Xiaomanlv was trapped on the cement floor.

In the case that the map route is not updated in time, there are two main reasons that Xiaomanlv failed to avoid obstacles:

First, in the daytime scene, neither the camera nor the LiDAR could identify the wet cement. Generally speaking, for automated logistics vehicles, this kind of ground should be judged as specular reflection and treated as a water surface. However, after the cement is 80% dry, there is no reflection, and this state is likely not to be judged as water.

Second, the wet cement scene belongs to the blind spot scene previously tested by logistics vehicle Xiaomanlv, similar to the long-tail scenarios. Long-tail scenarios refer to a wide variety of scenarios with a lower probability of occurrence or sudden occurrences, such as vehicles running red lights, pedestrians crossing the road, temporarily damaged traffic lights at intersections, and illegally parked vehicles on the roadside. These scenes are not frequently-occurred, circumstances are different or complicated, and difficult to handle. However, those scenes are one of the keys to the realization of autonomous driving technology.

Because the autonomous driving function is implemented based on AI algorithms. Based on current technology, AI algorithms can only perform tasks that have been trained before. Autonomous driving technology cannot simulate all complex road conditions. It can only approach infinitely all scenarios, but it is impossible to find all scenarios, so there are still many unsolved long-tail scenarios.

From a certain point of view, the occurrence of this new accident and new failure also helps to find new long-tail scenarios in automated logistics vehicles, and if those long-tail scenarios are solved, the speed of large-scale commercialization will be accelerated in logistics vehicles.

Large-scale commercial use: Though the road ahead is dangerous and difficult, we can only achieve our goals with constant efforts. 

As early as the end of 2015, Alibaba established the Cainiao ET logistics laboratory, focusing on the research and development of unmanned logistics vehicles.

In June 2018, Zhang Chunhui, then director of Alibaba Cainiao ET Logistics Laboratory, said: “In the next three years, Alibaba Cainiao unmanned devices will reach 100,000 units.”

In October 2021, on the occasion of the first anniversary of the launch of the automated logistics vehicle Xiaomanlv, Alibaba announced that Xiaomanlv has completed the “from 0 to 1” mass production, sold more than 350 units, and entered more than 200 universities and communities. But Alibaba’s goal three years ago is far from being achieved.

No matter how Alibaba plans to expand autonomous driving routes in the future, the low-speed and limited scenario of unmanned delivery will achieve large-scale commercial use more quickly.

lidar for logistics mobile robots

At this stage, Alibaba has achieved the layout of 350 units of Xiaomanlv, and is gradually approaching the latest plan of “expanding to 1,000 units in one year and deploying tens of thousands of units in three years”. Although there are still many difficulties to overcome in the operation of automated logistics vehicles, we firmly believe that the coexistence, co-creation, and win-win of the industrial chain will surely usher in a new development situation.

Neuvition S2 flash LiDAR

Titan S2 is the first flash LiDAR product in the Neuvition LiDAR family! S2 is specially designed for indoor close-range application scenarios mainly including logistics vehicles, supermarket flow statistics, privacy protection monitoring, security detection, indoor 3D imaging, volume measurement in trucks, industrial machine vision, AMR/AGV/RGV obstacle avoidance, etc.

Compared with similar LiDAR products based on pure solid-state LiDAR technology and the same-range detection on the market, the advantages of Titan S2 are as follows:
High resolution, long detection distance, high precision, large field of view, with camera

Neuvition S2 lidar point cloud