/diy/ - Do It Yourself

I've never researched robots whatsoever and I have zero knowledge on any electrical systems but I want to develop a robot that can follow me around and listen to my voice commands and do things I tell it to (eg. "Refill my drink" if I was sitting at a rsstatrunt and I didn't want to get up). Where would I start with this? Basically imagine Chappie in looks but less intelligent and probably less strong too

>>2653870
Could I make a screen for its face?

>>2653870
>Where would I start with this?
I've never gone snorkeling and in have zero experience swimming in a pool but I want to dive a wreck in the North Sea on heliox.

>>2653870
Start with a single board computer. Piece by piece you can learn the other parts of the robot to do what you describe.
Don't get discouraged by posts like >>2653886 , this is stuff that you can learn and do.

>>2653900
>discouraged
Learning to dive recreationally is similarly doable though a typical learning path.
OP want to build a bipedal platform with linguistic processing for goal, self-planning and processing, in a zero safety zone real-world application.
That is kiddie crayon drawing stuff.

>>2653930
So then tell them how to start diving recreationally. Early on, whatever they build probably won't be bipedal, probably won't be out in public acting completely autonomously. They would build what's doable, and it would work for them at home first. Nevertheless it's a step by step process and they are asking for the next few steps.

>>2653978
While I appreciate your desire to believe this is a mixed up kid who just needs a little paternal guidance to be set in the right path, the reality I've encountered time and again is lustful dreamers who assume everything is simple and have little to no ability to ingest time and attention toward achieving a goal.

>>2653996
>a little paternal guidance
But since it will be instructive to you, let's give it a shot.
>>2653870
Hey! If you're interested in learning robots, you'll really need to know a thing or two about control systems. Microcontrollers are a great and immediately useful way to get started! You can create tons of fun litttle wheeled robots with sensors out of them.
Here's a great, free course. You can use an emulator. But the investment in hardware is very small of you want to build for real. There's something very special about building a little robot that can do rudimentary path finding.
Good luck!
https://technobyte.org/8051-microcontroller-course/

>>2654000
Thanks, pardon the maybe optimistic mentality on my part but I stand by it, I appreciate you following through with good guidance.

>>2653835
>What projects are you working on?
I assume that a lot of people that have their robotics projects finished have papers published on their models, so, because of doxx risks, there won't be much showcasing to see

>>2653870
It's not feasible to use something like this in a restaurant because I don't think any restaurant owner would let you bring a big piece of machinery like that into their business, but I still think you should have a go with robotics
To build something like that you would need to brush up on quite a few fields, as previously mentioned
You'll need some AI knowledge for the speech recognition
Programming of course to tie all the control stuff together
Microcontrollers, as mentioned
Assembly things like soldering, manufacturing too, depending on size could be 3D printing or welding if you're picturing metal, CNC if you have access
There'd probably be a lot of 'support skills' you'd need too that will be the infrastructure for all these front and centre skills, CAD for example
Knowledge wise there's a lot of things you should know too, batteries, motors, sensors
You've already accepted that it's a lofty goal (probably completely unfeasible at this point in time to be honest) so I'm not saying this to dissuade you, just to give you an impression of how much there is in this topic (and hopefully that is exciting to you rather than upsetting)
>>2653886
>>2654000
Both of these guys are justified in what they're saying, depends on you as to which one is ultimately right

>>2654115
Applies to me
I'm about to start an interesting robotics project in academia and I can't talk to anyone online about it because it's so specific it instantly IDs me
And in real life I only really know 1 person who doesn't switch off as soon as they hear 'robot' from assuming it's way to complicated for them to understand LET ME PRACTICE MY LAYMAN'S ABSTRACT ON YOU

>>2654170
Most important but omitted: power systems.
The real reason these things don't exist is power. You can't carry enough on the platform to do anything useful for very long.
It's a special case of the tyranny-of-the-rocket-equation. More power needs heavier frame. Heavier frame needs bigger actuators. Bigger actuators need more power. Etc. Soon enough, you're building diesel engines driving generators on your mech, a.k.a. a locomotive.

>>2654170
>Both of these guys

>>2654180
Oh, my bad, I got you and OP mixed up when you replied to your previous post, I don't usually look at the name field
Good on you for giving the advice
>>2654178
Just spitballing here but I wonder if you could do a two stage type thing like rockets do to overcome that?
It would depend a lot on application but I can see a scenario where a big heavy diesel powered 'mother' robot could carry smaller and lower powered robots that do finer work to their destinations
I've not read up on the subject, maybe this is already a thing or wouldn't work

>>2654115
people publish papers on their line follower wheelbarrows?

>>2653870
it's called getting married

I take this as an invitation to advertise my channel
https://youtu.be/EJ_u63mdeNk
>>2653870
You can start by watching the video

>>2654170
There are restaurants in japan (Cafe Gasto) that have robot waiters to bring out the food.

>>2655712
I think there was one in America too, but they're typically on wheels rather than bipedal so less chance of falling on someone, if this is different though I will happily stand corrected
The main thing though is restaurants can make safety assessments of their own robots and weigh that against the publicity, but for this bipedal robot the guy is planning I don't see them ever being interested in taking the time to do that

>>2655708
Have you considered running the transcription, text gen and voice synthesis locally?
If you transcribe locally you would have a better shot at your detecting a word thing you mention in the video, because it could be constantly listening and just keep checking for that word, no need to upload to AWS
I think local text generation would be faster if you had even some middling hardware, and you can swap out models and have more context than 'Girl loves Boy', you can get more detailed so it knows your name and maybe something about what you like and so on
I don't really know anything about voice synthesis but I think they can usually run at 1:1 speed or better so again that might be faster than this cloud based way of doing things
The rentrys in the OP posts for /lmg/ and whatever the voice synth one is called can give you more info if you're interested, don't bother reading the threads themselves though

final.sh:76 why do you assign a random number to the recording?

>>2655869
For text to speech there is espeak but it sounds like microsoft sam, for speech to text however I don't know if there is any good ones. I will howver might have to look into it because its pretty slow. Some anon at alogs.space however did give me some code to listen to the volume so I won't have to press the button. But its in C++ and its kind of cryptic the way he wrote it.

>>2655892
post the code, I will see if I can explain or maybe someone else can
https://www.youtube.com/watch?v=c_0Q3T0XYTA
this is the top result for 'speech-to-text' on github, looks fine, not checked any further than a 5 minute search though
you'll need speech-to-text to make sure it's the right word at high volume anyway so might as well just use speech-to-text on its own and forget about the volume consideration

>>2655894
Here is the pastebin
https://pastebin.com/aBF0fEyv
>I set it to expire after 1 week
Anyways it works great, it calibrates the volume and records when you speak into the mic
>if you set the calibration to 10 instead of 2.5
But
It keeps going after its finished recording and i want it to stop after its done recording. I could have asked the anon but I don't want to keep pestering him.

does something like working with hand tools even have a place in robotics or is it just high level skill like operating machinery, programming that are useful

>>2656947
I'd say it depends on what you're doing
Do you need to build a frame for it? Do you have something like a 3D printer for a shell or will you have to make it yourself? Does it need to be bolted down to something if it's non-mobile but deals with forces?
A soldering iron is a given for this stuff and I think the rest is 'maybe'

>>2655892
There's low footprint AI to do speech to text and text to speech, Piper and Whisper.

Anyone have experience with GTSAM and tuning noisemodels?

I already have reasonably good cloud registration, and get can get coherent maps, but I seemingly get a lot of angular drift.

what do you guys think is a good robot aesthetic?
I think some of the metal gears look pretty cool, and so do the KUKA industrial arms, the boxguard from deus ex is nice too
I don't really like those smooth white and blue plastic ones though, they look like toys to me

>>2659031
forgot pic

>>2653871
the screen would be a nice starting point for you, you can get a matrix screen and have it make different expressions based on a sensor of your choice

This internet-controllable robot is online now as well: https://remocon.tv/645f7fc079e18b0016da549e

>>2660339
just had a go, that was pretty fun, did a few laps of your basement then once under the stairs