A high school up master created a lazy robot in 200 days! It can carry bags and follow you in all scenes. It has been played millions of times

Bowen Xiaoxiao sent from Aofei Temple
Quantum Bit Report | Public Account QbitAI

It’s hard to imagine that this robot was made independently by an 18-year-old high school student !

Tracking is possible both indoors and outdoors:

Even if I follow you in the dark, it’s no problem:

It's totally fine to be a mount:

All the work, including principle determination, appearance design, circuit connection and code writing, took 200 days.

Moreover, all of this was done by one person , a B-station Up master named Ling Shiqi !

The video received nearly one million views in less than three days after being uploaded.

Originally, the independently directed robot story and the entire process of hardcore DIY robot are already eye-catching enough.

At the end of the video, the Up master also took out his musical instrument and dedicated a song to this project, which directly triggered the barrage:

Is there anything else you don’t know how to do?

Hexagon Warriors!

Treasure boy?!

From programmer, to director, to music lover, switch between multiple identities seamlessly!

What kind of Up Master is this?

What kind of robot is this?

From modeling to production, fully independent design

What kind of robot do you want to make?

Ling Shiqi gave the standard in his mind:

Steel structure, plastic shell.

It can follow, run, carry things, be pulled, and most importantly... it can sit.

After drawing the design, the 18-year-old boy pondered and said: It's feasible.

Let’s start, step one: modeling !

Let's take a look at the modeling workload:

Looking closer, the design of the entire body is not a rough and straightforward geometric shape, but is designed with reference to ergonomics, from the aspects of stability and practicality:

After the modeling is completed, the next step is to make the head bald as much as the modeling.

Shell production .

To bring the model from the computer to reality, Ling Shiqi chose water-washable resin as the material for 3D modeling.

This doesn't sound like much work, just sit in front of the machine and wait for the finished product:

But when you actually try it, there are too many problems:

• Using water-washable resin as printing consumables may cause cracking problems

• When the finished shell is too thin, more severe deformation will occur, which will make it impossible to install

• Oversized parts require very high assembly accuracy

There is no shortcut in this process, only repeated attempts, adjustments, attempts, and adjustments.

The shape is formed, and the next step is to put filler on the parts .

This process can unify the base color while filling in the unevenness of the printed surface.

Then comes the final step of the exterior production, painting:

Now, the "very unskilled, very boring and time-consuming" shell production process is completed.

Let’s look at the robot’s eyes. Ling Shiqi used a 2D laser radar :

He used two radars in total and rewrote the unpacking algorithms according to the communication rules.

Okay, the next step is to make the robot’s heart!

Ling Shiqi chose a single-chip microcomputer as the main control device to control the brushless motor and servo, and to make the final judgment on remote control, driving and following speed direction:

Then use the Raspberry Pi as the tracking algorithm output device and transmit the data to the microcontroller via the serial port:

The MCU and Raspberry Pi here are both powered by a unified 10A step-down module .

However, since the current that the Raspberry Pi USB can provide is very small, and the radar is powered directly from the USB port, it often falls into an infinite restart state at the beginning.

Therefore, Ling Shiqi solved this problem by connecting the circuit supply voltage (VCC) in the radar to 5V separately.

As for the complete tracking function, Ling Shiqi used a total of 250 lines of code to implement it.

Compared with shell production, code writing can be described as "highly technical, very non-boring and time-consuming".

After all, the last list of solved problems has the most examples of programs:

To be able to track targets beyond the hardware supported range, the tracking angle is limited, and the target is searched again after exceeding the range;

To prevent the robot from running around when it loses the target, it is programmed to stop tracking when it detects that the target is too far away.

It must also be able to communicate with the microcontroller in a timely manner to prevent data loss due to excessive communication, as well as nonlinear steering algorithms...

In the end, the functions mentioned in the design drawings have been basically realized.

In addition to the features mentioned at the beginning, storing things is also completely fine:

Standing on his own very convincing messy ground, Ling Shiqi said:

This is the most difficult project I have ever developed in my life.

The story behind the robot

Yes, building a robot from 0 to 1 may not be difficult for students majoring in robotics, but for high school students, they need to learn all the technical knowledge from programming, circuits to automatic control by themselves .

In terms of software, all programs and hardware and software codes need to be debugged and optimized by ourselves, and the number of bugs fixed in the algorithm can be said to be "countless":

In terms of hardware, when modeling the fuselage, Ling Shiqi specifically learned the relevant design methods:

But when I did it myself, it took me a long time just to design the shell, which seemed to be a "non-technical" job, and I broke several shells:

It is conceivable that in addition to the shell, there are many parts that need to be replaced in modeling and design.

Ling Shiqi himself said that the final waste of spare parts can even "build another robot":

After finally completing it, the robot was damaged again.

The design began in August 2020 and production began in October, but during outdoor shooting in December, the robot's "head" accidentally fell off.

Since the servo at the neck connection was not very stable, it was initially just attached to the robot and could be easily removed. As a result, although there were no problems during the home experiment, it could not hold on on uneven roads and eventually broke.

After learning about this situation, Ling Shiqi improved the robot's structure and fixed everything with screws, and similar problems never occurred again.

But because of this incident, the project was forced to be suspended for a period of time. Later, nearly 1/3 of the robot shell was dismantled and remade, and it was completed around April 2021.

Of course, it’s not just the outer shell that has been adjusted, the software has also been upgraded in this modification.

Three or four algorithms have been added to the radar, mainly to optimize invalid data of communication and radar, and several others to eliminate possible misleading during tracking or some situations that seriously affect the operation effect;

The driving algorithm has also undergone parameter adjustment and optimization, but this operation is not standardized and is basically trial and error. If a more formal project is carried out, a formal model may be calculated before programming.

All algorithms were written by Ling Shiqi himself without reference to other algorithms:

After all, I had never seen a similar product, and when I determined the principle, I had already thought of the logic in my mind, so I didn’t think about looking for other people’s code.

After 9 months of polishing, the robot is excellent in all aspects, but for Ling Shiqi, there is still room for improvement.

On the one hand, robots cannot make facial expressions, and all expressions in the videos are added through post-processing, which is a special effects production.

Ling Shiqi said that he originally wanted to outsource the special effects, but found that the outsourcing capacity was insufficient and the special effects produced were too lame... So he learned to do it himself, and the effect was indeed better.

Of course, it is also possible to display facial expressions directly, but "there are already very mature algorithms on the market", so he did not consider it as a "highlight" of the robot.

On the other hand, the robot itself has poor physical stability; the two generations of lidar used in the obstacle avoidance algorithm are also affected by lighting, the results are not as expected, and the outdoor performance is not good enough.

If there is time for further optimization, Ling Shiqi plans to combine two lidars and cameras to make outdoor tracking more stable.

As for why to make this robot?

Ling Shiqi said that he initially wanted to find someone to help carry the equipment (such as the tripod and camera used to record the video) , but later realized that this task could be "done without human intervention."

Wouldn’t it be great if we could make an automatic truck that could also serve as a driver?

Although a lot of effort was put into the later stage, the effect is still very good in terms of completion.

The name of the robot Devil comes from a song called "Devil Swing (Remix)".

It was also because of this project that Ling Shiqi decided to change his major from Electronic Engineering (he had been admitted to Georgia Institute of Technology) to study AI and Robotics .

△ The major you originally wanted to study

In an interview with Quantum位, Ling Shiqi said that through this project he realized that algorithms might be the dominant force in the future, so he wanted to focus more on programming.

“I want to change my major to AI & Robotics”

Ling Shiqi is from the Shanghai Normal University Affiliated Middle School (SNU Affiliated Middle School). Like other middle school students, he spent his first 16 years "basically preparing for exams".

In his first year of high school, Ling Shiqi came into contact with robots through a competition, and from then on opened the door to a new world.

As can be seen in the video, in addition to winning first prizes in the "Robot Comprehensive Skills" and "Electronic Control Project" competitions, Ling Shiqi also invented a "remote control device for special purposes of laser positioning" and won multiple awards.

Ling Shiqi said that although these competitions were not of high quality, they successfully aroused his interest in robots:

When I first discovered that I could make things outside of myself run according to my will, I was so moved and shocked, and everything felt like it had never been so meaningful.

Robots are the first light that illuminates my life path.

He started uploading videos in 2019, and in early 2020 he made a knife with a circuit that glowed. The number of views jumped from tens of thousands to more than one million:

In the next video, the knife evolved directly into a lightsaber that could spit fire and change color. The number of views reached over 2 million, which also confirmed the reputation of the "handmade UP master":

After that, it took 9 months to build the robot Xiao D, and it was 2021 in the blink of an eye.

During this period, Ling Shiqi once had the idea of ​​giving up, "I felt like I would become abnormal if I continued", but he finally persevered.

It was also in this project that he found his future direction.

Many big Vs enthusiastically forwarded this robot:

Station B officials also warmly encouraged:

It is worth mentioning that Ling Shiqi’s skills are not limited to robotics, directing and piano playing. According to his high school classmates, skills in painting, composing, etc. have also been unlocked.

As for the future, Ling Shiqi said that he is more likely to start a business, make products or engage in creation. If he devotes himself to scientific research, he will most likely want to study the direction of general artificial intelligence (AGI) .

Although this robot does not yet involve AI algorithms, Ling Shiqi said that he has planned to start learning from CV and do some related AI small projects, such as object and shape recognition algorithms.

What kind of sparks will he create in the field of AI?

let us wait and see.

Video address:
https://www.bilibili.com/video/BV1s44y1173e