'Self-driving bicycle' that goes swiftly unmanned is born, developed by Huawei engineers



With the evolution of sensors, cameras, and AI technology, the technology of 'autonomous driving systems' in which computers drive cars instead of humans is also advancing day by day. And was able to implement such automatic operation system to a whopping two-wheel traveling bicycle, at the age of 18

electronic University of Science and Technology and graduated from, working as AI engineer of Huawei Itokena晖Mr. has been reported in the movie.

[Self-control] Self-guided bicycle completed Autonomous bicycle! !! [Hardcore] --YouTube


Mr. Wakaaki developed the self-driving bicycle because he himself rode a bicycle on a rainy day and fell and injured his face.



The point of self-driving cars is that 'cars can drive even if the driver's seat is unmanned.' On the other hand, in the case of self-driving bicycles, there is a problem in that 'if the driver's seat is unmanned, you cannot stand on your own.' A typical car has four wheels, so it can stand on its own, but in the case of a bicycle with only two wheels, a human rides and moves the center of gravity to maintain balance. If you are completely unmanned, you will not be able to stand on your own unless you reach a certain speed.



So, Mr. Wakaaki wants to implement an automatic driving system on his bicycle ...



We have adopted an attitude control module that uses the rotational moment of the disk.



It is said that the same attitude control module is also applied to artificial satellites.



In self-driving bicycles, a small, high-precision sensor detects tilt and feeds back the data to the attitude control module so that it can stand on its own.



In addition, to enable autonomous driving, the top tube is equipped with a

LIDAR sensor that recognizes surrounding objects with a pulse laser.



The handlebar is equipped with a camera for image recognition.



In addition, we have adopted a system that rotates the rear wheels directly with a motor without pedaling.



Design and create the necessary parts with CAD.



The rear wheel drive module installed in the seat tube looks like this.



Housed above the rear-wheel drive module is a computer board for autonomous driving.



This board was designed by Mr. Wakaaki, who has a hobby of electronic work.



The processor for AI is Huawei's

Ascend 310 .



This is the board that was actually manufactured. The AI for autonomous driving was originally developed by Mr. Wakaaki using

Huawei's Application Orchestration Service.



In addition, the handle is a mechanism that moves using a motor and wire attached to the top tube.



The completed self-driving bicycle looks like this.



After testing the balancer in my room ...



We will actually test it outdoors.



When Mr. Wakaaki raises the bicycle stand ...



A bicycle that no one is straddling is perfectly independent on the spot.




Even if Mr. Wakaaki pushes the bicycle with his finger, there is no sign of falling.



Even if you hang a paper bag with luggage on one side of the handle ...



The balance is not lost at all.



The balance control by the attitude control module is perfect, and it can stand on a fence that is only a few centimeters wide without any problems and does not make slight movements.



The attitude control module can also be fixed to the top tube ...



You can also raise the front wheels to maintain balance in a unicycle-like state.



And the test of automatic driving.



The bike is slow, but it's running fine. It also recognizes traffic signs drawn in the lane, and when you turn right, drive in the right turn lane.



Since the camera recognizes the object in front, obstacles on the road and ...



While recognizing the car parked on the shoulder, it automatically takes the course and runs.


Since LIDAR can recognize the surrounding walls, the bicycle can go smoothly even in narrow corridors.



The movie also demonstrates back driving on slopes. In addition, Mr. Wakaaki named this bicycle automatic driving system Extremely Unnatural Auto-Navigation (XUAN), and the source code is published on

GitHub.

in Hardware,   Ride,   Video, Posted by log1i_yk