跟读练习: I Built a Fully Autonomous Flying Umbrella - 通过视频学习英语口语
加载中...
1
This is the flying umbrella and today we're going to make it follow me around.
2
This is version 2 of my flying umbrella project.
3
Version 1 was built all the way back in...
4
2024. I basically just slapped some propellers onto an umbrella and it somehow managed to fly.
5
Except, as thousands of you guys pointed out, it doesn't follow you.
6
What if you made the drone follow you instead of using a controller?
7
Make the umbrella follow you.
8
Keen to see you follow you automatically.
9
So today, we're finally going to make things right.
10
Unfortunately, the current state of the old umbrella is, well...
11
So we're gonna redesign everything.
12
More reliable, more convenient, and most importantly, fully autonomous.
13
I mean, what could go wrong?
14
If you're wondering what this is, let me explain my plan.
15
Before we build the upgraded flying umbrella, we need a tracking system that can follow a person.
16
This can be done with a camera mounted under the umbrella, a depth sensor or lidar, a GPS, or even overseas teleoperators.
17
No matter which method we use, it's going to take a lot of testing.
18
Doing all that testing on a giant frame like the umbrella would be a nightmare.
19
So instead, I'm building this smaller, normal-looking drone to use as a testing platform.
20
There was no reason to design a custom drone just for testing, so I reached out to Hollibro and they sent me this frame.
21
To control and stabilize the drone, they also sent me this professional-grade flight controller.
22
After plugging in the motors, GPS, and receiver, I was ready to start setting up the flight software.
23
Apparently, it's super easy and should only take a few hours.
24
It's currently day 16, and that took way longer than it should have.
25
Although I did run into a few issues here and there, such as these misleading diagrams, the main reason it took so long is because I procrastinated.
26
Hopefully this doesn't happen again for the rest of the project.
27
Okay, great.
28
Now we have a reliable testing drone, but in order for it to autonomously follow a person around, we're going to need a few more components. So...
29
This is a camera module, and this is another flight controller.
30
But this time, we have a slot inside for a Raspberry Pi.
31
The Raspberry Pi's job is to process the camera footage and pinpoint where my head is in real time.
32
Then, it tells the flight controller to fly the drone towards this point so that it constantly stays centered.
33
This should be pretty simple, but I only have two weeks to work on this project before school starts, so we're going to get some help.
34
This is Hinson.
35
He studies computer science at a relatively small university called Stanford, and he's one of the most cracked programmers I know.
36
While Hinson starts working on the tracking code, I'll go and start building the actual umbrella.
37
Now, to make an umbrella fly, you need to put propellers on the sides.
38
But the only solid enough mounting point is the middle rod.
39
Yeah, do you have a solution?
40
Yeah, what if we just connected the four propellers to the center using some sort of big frame?
41
Well, we actually tried that last time, but the giant frame was just way too cumbersome and annoying to bring around.
42
Hmm, I guess you are right.
43
if only if they're...
44
Wait a second.
45
What if we did folding arms?
46
I feel like this mechanism shouldn't be that hard to recreate.
47
Let's make a quick test model as a proof of concept.
48
Nice.
49
We also need the arms to lock firmly into place upon opening, but it seems like this mechanism has way too much wiggle room.
50
But then I realized by simply changing the axes of rotation, for this locking plate, we get rid of this problem completely.
51
After a lot more designing, printing, and testing, I was able to come up with this design.
52
Now let's just put four of them together and create our internal frame.
53
If you're wondering how I'm able to use CAD directly in my browser, it's because I'm using Onshape.
54
You can try it out six months for free using the link in my description.
55
Okay, so here's the final design.
56
We have the central hub that attaches to the umbrella, the locking mechanisms, the hinges, and these arms that we can simply reuse from last time,
57
although I first had to re-solder some of the wires and add some better ESCs.
58
Let's get the rest of these parts 3D printed.
59
Oh shi- At the start of this project,
60
I was using my BambooLab X1C, but recently, BambooLab sent me their new H2D.
61
These hinges were designed to be printed out of carbon fiber nylon for maximum strength, but doing so used to be a huge pain.
62
On the H2D, however, you can simply dry the filament in the AMSHT, and printing is as easy as pressing a button and watching it go.
63
Thanks to its fully heated chamber and advanced filter, tough engineering filmments like CF Nylon now print without any warping or smell.
64
The quality of these prints is literally mind-blowing.
65
If I had this earlier, this project would have been so much easier.
66
The central hub-wee design looks very cool, but with all these gaps in weird geometries, it required a ton of supports, which left rough surfaces upon removal.
67
But the H2D has two nozzles that can swap within just a few seconds.
68
This means you can print in multiple materials and in our case use support filament with minimal changes to print time.
69
For advanced engineering projects, the H2D is one of the best 3D printers you can get.
70
But if you're just getting started, Bamboo Lab has options for every budget and skill level.
71
To learn more, check out the link in the description.
72
Thank you Bamboo Lab for supporting my projects and sponsoring this video.
73
These supports remove so much cleaner than the previous ones.
74
Now let's assemble everything.
75
Originally I wanted to attach the umbrella to the frame permanently and create some sort of fancy collapsing mechanism, but we're just gonna friction fit it because I got too lazy.
76
This rod sticks out a bit too much, so let's cut that off.
77
Somehow the umbrella still works normally?
78
Let's finish assembling the frame.
79
All these little locking mechanisms are also printed with CF nylon and slid into place.
80
And now we can just lock them in with some screws.
81
Then I had to add 16 more small screws to attach rubber bands.
82
I couldn't fit my screwdriver in the small gap, so I had to use this Allen wrench over and over which took forever.
83
But now we can finally start attaching the arms and of course do some great cable management.
84
Now, when we fold the arm outwards, we can watch the rubber bands engage the mechanism and lock the arm in place.
85
Nice.
86
Isn't that actually crazy?
87
And now it's like the size of a tripod.
88
And I just go boom, boom, boom.
89
How did I even make this?
90
Just like the testing drone, let's add the GPS, the flight controller, the receiver, and other various electronics.
91
And now, we're ready for our first test flight.
92
Okay, flight test one.
93
Would the complex folding arms hold up, or would they cause too many vibrations for the flight controller?
94
We were about to find out.
95
No clue if this will work, first flight.
96
It seemed like the propellers were spinning the wrong way, but since I had my laptop, I was able to quickly reverse the motor directions and...
97
The drone was still pushing itself into the ground, so I did a bit of troubleshooting and turns out...
98
I can't reverse them in software.
99
Let's just do this tomorrow, it's too dark anyways.
100
Okay.
101
Wow, it's tomorrow ready?
102
That was fast.
103
I actually had to rewire and resolder all the motors at 12am which took forever but now… it finally…
104
Maybe it's just a fluke, let's try that again.
105
After a bit of troubleshooting, I was able to fix the problem and now… It was a bit shaky at first,
106
but after using the GPS to set up a position hold, I was able to get this.
107
And just for the fun of it, I decided to try attaching the umbrella as well.
108
But let's save the full reveal for when the autonomous tracking feature is done.
109
Speaking of that, the camera we chose earlier requires a custom tracking code and doesn't work if it's too dark.
110
So instead, we're going to use… Wait John, hear me at… This isn't your ordinary camera.
111
This is a time of flight camera and it sees in depth.
112
Unlike a normal camera that captures light, this one emits light.
113
Let's imagine the simplest scenario, trying to detect the distance to that wall.
114
The top camera emits a single wave of light, which bounces off the wall and comes back.
115
Now if you look closely, the returning light looks almost the same as what was sent out, just slightly shifted.
116
This is called the phase shift, and by measuring it, the camera can directly calculate the distance.
117
Isn't that just LiDAR?
118
You might be thinking, isn't that just lighter?
119
Most lighter use direct time of flight, which times how long a light pulse takes to return, which works great for these large buildings.
120
But at short distances, those nanosecond differences become super hard to measure.
121
By using an indirect time of flight camera, we measure how long the light shifts, which works better for short distances and allows the camera to be much more compact.
122
But what if you're aiming at something that isn't just a flat wall?
123
Well, instead of sending a single beam of light, the camera sends light in all directions.
124
Since the distance is directly proportional to the phase shift, we get this beautiful depth picture.
125
Wow, that's pretty cool.
126
Now, in theory, using this camera is a pretty good choice, but in reality...
127
Really, what does this even mean?
128
We got an ArduCam Uncone error.
129
Oh, would you look at that?
130
There's absolutely no description of what an Uncone error is.
131
By some miraculous miracle, and by downloading a bunch of firmware that we probably weren't supposed to, even the time-of-flight camera was now starting to work,
132
and the tracking code was coming along together as well.
133
Now, it was only a matter of integrating everything together.
134
That's not supposed to happen.
135
Okay, so we had just got the top camera working, and I decided to hold it high up above my head to test it out.
136
But as I was bringing my hand up, it disconnected.
137
And from that moment on, this board just refused to reconnect.
138
No matter who we asked, no matter what we tried, this thing was just completely bricked.
139
Is it the Raspberry Pi that's broken?
140
Or maybe it's the carrier board?
141
Or maybe it's the flight controller?
142
Or maybe the cable broke?
143
Or maybe my laptop isn't working?
144
After a lot of troubleshooting, we determined it was probably the Raspberry Pi.
145
But even if we were correct, with only 6 days left and the 4 day shipping time, we were cutting it really close.
146
While waiting for the Raspberry Pi to arrive, I took the Bear Frame out for a few more test flights to perform some PID tuning
147
so that it could fly more smoothly.
148
But as the days passed, I couldn't stop worrying about how this project would turn out.
149
Dude, that's horrendous.
150
Yeah?
151
I think I watched the video.
152
Wait, actually?
153
Yeah, that was different.
154
I guess that's our project number two.
155
Little did he know, this alleged project number two was on the brink of collapse.
156
If this doesn't work, the project is officially over.
157
Thank you guys.
158
In three, two, one.
159
Oh my god.
160
Uh, well it did something different.
161
Oh my god.
162
I'm pretty sure.
163
Wait, I think it worked.
164
It did not work.
165
No water, no water.
166
And just like
167
that, the project was officially over.
168
I mean, we both had to go to school and we just couldn't afford to dedicate any more time.
169
But no matter how busy I got, I just couldn't stop thinking about the project.
170
So, one day, I decided to give things another go.
171
This time we're replacing every single component with a brand new one, so there should be no way anything fails to connect again.
172
Let's go, it finally works again!
173
Everything was going well at first, but when I tried to connect the only component I didn't replace, the time of flight camera, that didn't work.
174
So I decided to scrap everything and try using a GPS system instead.
175
So now this looks like a bop- It actually worked pretty well, but the GPS was only accurate to about 3 meters,
176
and getting one with enough precision would cost thousands of dollars.
177
It's just like, not that good.
178
So once again, I gave up.
179
It wasn't until half a year later that I decided to give this thing one final try.
180
This time, I decided to bite the bullet and buy another top camera.
181
Oh yeah, and Hinson's back from school for the summer nap.
182
Maybe the real flying umbrella was just the friends we made along the way.
183
Despite a lot more trial and error and a lot more struggles.
184
30 clips of me walking into this field, and then 30 more clips of me walking back with absolutely zero results.
185
Eventually, we got to the point where we were finally able
186
to move from the testing drone to the full-on umbrella and test this thing out.
187
As I started transferring all the tracking hardware from the tracking drone to the umbrella, I honestly had no idea what to expect.
188
What was supposed to be a simple project ended up coming with so many unexpected and unlucky moments.
189
In fact, I barely scratched the surface of what we went through.
190
You can pause here for the full list.
191
Yet somehow, we managed to push through, and tonight, an entire 358 days since the start of this project,
192
we were now attempting to bring to life an idea that had first stemmed three years ago.
193
If this still didn't work, I honestly didn't know what I was going to do.
194
But sometimes you just need to ignore all the doubts in your mind and go for it.
195
Even as it started getting darker, we just couldn't stop playing with this thing.
196
Sir, this is what a flying umbrella is in 2025.
197
Who wants to hold that TV?
198
Let's go!
199
Now, I would be lying if I said that this project turned out perfect.
200
But at the same time, I would also be lying to say that this moment didn't bring us joy.
201
I mean, just look at it.
202
It's like a dream.
203
A literal floating umbrella.
204
Just dancing in the air.
205
I've been a perfectionist my entire life.
206
When we started this project, every single detail mattered.
207
So when things went wrong or weren't perfect, it always felt crushing.
208
But over the past year, I've learned that you can't let perfectionism stop you from finishing.
209
Sometimes, it's not about making something perfect.
210
It's about making it real.
211
And once it was finally real, there was only one thing left to test.
212
That's right, this thing even works in heavy rain.
关于本课
您正在使用跟读技巧通过视频"I Built a Fully Autonomous Flying Umbrella"练习英语口语和发音。
每天练习15到30分钟，将显著提高您的英语流利度和发音准确度。
什么是跟读法？
跟读法 (Shadowing) 是一种有科学依据的语言学习技巧，最初开发用于专业口译员的培训，并由多语言者Alexander Arguelles博士普及。这个方法简单而强大：您在听英语母语原声的同时立即大声重复——就像是一个延迟1-2秒紧跟说话者的影子。与被动听力或语法练习不同，跟读法强迫您的大脑和口腔肌肉同时处理并模仿真实的讲话模式。研究表明它能显着提高发音准确性，语调，节奏，连读，听力理解和口语流利度——使其成为雅思口语备考和真实英语交流最有效的方法之一。