/sci/ - Science & Math

>>7107593

I started my own robotics company with an engineer after my CS degree so that's how I got into the field.

>>7107593
>I'm working on BS in Comp Sci

I am so sorry.

>>7107594
Oh that's awesome, I'm assuming you teamed up with a Mech guy?

>>7107593
>BS, not BA
Have fun programming soulless automatons.

>>7107598

actually it was an optics guy but it worked out.

>>7107593
why comp sci then robotics? just why? you could have gone with CE or even EE and it would prepare you for robotics way way way more. for fuck's sake there is nothing in comp sci even related to robotics. drop out start again

>>7107604
how exactly did this come about? Also, what do you guys do today? Assuming that the company is still active.

>>7107617

did you fucking study comp sci?

no? well, then shut the fuck up.

here are a couple things you learn in comp sci:

system architecture
system modeling
software engineering
machine learning
data mining
data structures
signal processing
automata design
various algorithms including pathfinding
artificial intelligence

and many, many more things very fucking useful for robotics.

>>7107617
my school doesn't offer any engineering, and I'm not in the financial position to go to another school. I'm working with the best I can manage.

>>7107623

we knew each other a long way back.

eventually I started to get serious and brought him on board to do just general things together - 3d modeling, design, prototyping, etc. we're currently looking to get tech grants to accelerate the project a little and get a merchantable product out there sooner and hire more employees, among which there will hopefully be enough money left for an accountant. I hate doing that shit.

in terms of work we're currently still concerned with revising the prototype. it's a lot of measuring and prototyping and reiterating and learning how the various technologies work. It's hard, painstaking work and I can see why no company in our niche is advancing very fast, but I expect the rewards to be worth it once we're over the initial hump and start applying our edge.

>>7107624
have you stepped foot into a robotics lab?

>>7107624
>did you fucking study comp sci?

Yes as well as CE and CS is fucking trivial in comparison.

i'm in a similar position, comp sci for robotics is going to be high level control stuff, i've consigned myself to basically learning CE out of college and through work experience "working down"

>>7107639
care to share any hints as to what sector you're in? i'm curious now
also, invest $1000 into a good hobby level 3d printer for design iterations and prototyping

>>7107639
That's pretty awesome anon, hopefully you guys get your work completed. Best of luck to you guys. I've got a couple engineering buddies too, maybe I'll have to talk to them.

>>7107641
>no reply

>>7107649

so you have an undergrad degree in CS and an undergrad degree in CE

good job.

>>7107641
I imagine that indeed, you will need to learn some new things and theory is not praxis, but you can't sell me that there is no use for theoretical CS - unless you're talking about industrial robotics - but even then your quality control will be quite sub-par without intelligent systems.

I'm not discounting CE, there is definitely a place for CE, and a CE can learn all the things that a CS could do, but I do think a CS is better equipped to deal with these systems on a higher level of abstraction and functionality. I think a CE on the other hand would be better equipped to provide the interface between the hardware and the higher level software.

>>7107624
you are just a little babi scientist, you didn't make anything doing comp sci and you know it, because you don't actually make anything doing comp sci, you just learn a bit of theory that you will NEVER use because you WILL end up being just another code monkey like every comp sci student. I'm sorry but that's the reality of the situation. Womyn studies were better for robotics than comp sci is

>>7107658

we're working on atlas types.

>$1000 into a good hobby level 3d printer for design iterations and prototyping

I'm a million steps ahead of you buddy XD. we already have a small lab with all the bells and whistles. the only commodity we are really short on is manpower.

but this is an interesting phenomenon. when you do something novel and talk to people about it they'll be interested in sharing their ideas about how your stuff could or should be done. "electromagnets man", "hydraulics man" or "3d printers man".

>>7107667
thanks! and good luck to you too. I really hope it works out for you as well. With the economy in the shitter the best thing you can do is start a business I think.

>>7107677
> no use for theoretical CS
there isn't.
I know of two CS guys who worked in our lab.
They spent most of their time grinding and filing burrs.

If you want to work on physical robots, the best course is to go EE.

> CS taking signal processing
at what school?

>>7107691

you might be surprised that this trolling is actually sort of relevant. If you apply for tech grants your proposals, after technical expert review panels and approvals, get passed on to an ethics panel that decides whether your shit is "socially kosher" I like to say. Those are full of these womyn studies and sociologists. So yeh, your ironic shitposting isn't actually full on wrong.

>>7107707
How much did you spend on your 3D printer?
Ours was a couple thousand and it's incredibly hard to work with.

>>7107691
>do CS
>graduate and get job as DBA
>barely have to code
>literally do jack shit and get paid for it as long at things don't break

I used to feel for losers like you who failed to become anything, but then I remember how big of faggots you are and how you deserve what you got.

>>7107715

obviously you're not completely wrong. I can't give you any realistic numbers but you only really need about one computer scientist for every 100 engineers or so. but really it's like that everywhere - not just in robotics. You don't *need* more than one computer scientist in any project. You might want two so that they can work more efficiently, but a third one would be counterproductive again.

>at what school?

every half decent school that offers medical computer science

>>7107719

approx. 1000 for a FDM. we're thinking about getting two more, an SLA and a large volume FDM.

Let's say there's definitely a learning curve: the plastic expands and contracts, and the printer is a mechanical system that requires calibration. For instance we noticed that holes were always too small and extrusions were always too big, and sometimes the pieces warped as they cooled, giving inconsistent results. So now when the print is done and still hot, we flash-freeze it to -50°C which works out really well. the refrigerant permeates the hot and supple filament and cools it consistently, doing away with the warps.

>>7107707
>atlas types
interesting, i thought most of the really necessary bits for atlas types (mostly bipedal walking and posture) were heavily bound up in other proprietary projects?

are you guys reinventing the wheel?

furthermore, what's the application? military? search and rescue? medical assistance in hospitals?

>>7107747
you do realize the way to prevent warping is to print in a heated, enclosed build chamber right?

>>7107707
You said you started your own company. How did you get the money? Also, did you start straight out of uni or what?

>>7107807
>are you guys reinventing the wheel?
partly yes. reinvention is complete. now it just needs legs (huehue). If we see it complete it'll make darpa look like an amateur high school project.

>furthermore, what's the application? military? search and rescue? medical assistance in hospitals?

all of the above and more (maybe in the future). initially we're starting with radiation applications (because that's where I have contacts). it'll reduce and in some cases even eliminate downtimes in nuclear power plants and high energy research facilities (eg. CERN). However, if we get a new funding source we might move in a different direction because radiation robotics is a whole other type of beast.

>>7107812

the warping commonly happens during the cooling process because hot air gets trapped in the lattice, causing an insulating effect - the exterior cools much faster and contracts. so there's a shear force on the layers, pulling them apart. The ice treatment makes the plastic a little more brittle (why, we don't know yet), but there's better surface adhesion and less warping. just our observation. YMMV because every setup is slightly different, but we learned to work with ours.

>>7107814
Started during my final year. I had a little money. plus a loan. but it's a frightening experience. like if I flop, I could be homeless. but if I don't try, I'd be working 9-5 in some bullshit. if shit goes really far south I'll declare bankruptcy and move in with my parents. At least I can say I tried.

>>7107823
You know each ATLAS robot cost millions of dollars each.

>>7107593
Look OP I really want to tell you to just go find a job at some robotics company because that way you wouldn't compete with me, but I can't bring myself to do that.

Truth is you need to go to grad school before you can do anything cool

>>7107747
>>So now when the print is done and still hot, we flash-freeze it to -50°C which works out really well. the refrigerant permeates the hot and supple filament and cools it consistently, doing away with the warps.
if that really works well then you should patent it before some one else does, it may not be practical, but it is novel. You might run into some interesting stress relaxation issues over time though.

>>7107807
>>really necessary bits for atlas types (mostly bipedal walking and posture) were heavily bound up in other proprietary projects?
Boston Dynamics is a fucking asshole that doesn't publish anything.

Walking is actually sort of simple if you have force controlled actuators, but it still ain't a cakewalk.

>>7107823
>>If we see it complete it'll make darpa look like an amateur high school project.
I highly doubt that. Whole body control, or coordinating both the arms and the legs so you can perform a task is a pretty damn hard problem.

>>radiation
how are you going to rad-harden everything? There are some big uncertainties about radiation hardness of robots:
https://ece.osu.edu/news/2012/11/dtra-funds-study-radiation-hardened-robotics

Now tell me at least one thing, what are you using for actuators?

>>7107831
how can i contribute?

>>7109006
I too would like to help OP. My skills include eating my own semen and having absolutely no shame about living in my mom's basement.

>>7107593
1. arduino
2. ???
3. profit

>>7107847
>You know each ATLAS robot cost millions of dollars each.

indeed. but not for much longer. (~20 years)

>>7108976
>you should patent it before some one else does, it may not be practical, but it is novel.

na, there's prior art on this. but it's not patented as far as I'm aware and if it will be, we will have prior art on it too.

>>I highly doubt that. Whole body control, or coordinating both the arms and the legs so you can perform a task is a pretty damn hard problem.

I am proud to say that we have solved that problem.

>>how are you going to rad-harden everything?

we have a solution for this too. it involves a more intelligent approach than radiation-hardening everything. this platform should be able to operate in environments where non-browning lenses brown. this is partly thanks to recent advances in software and materials science and I think it's pretty neat.

>>7109006
If you're willing to work for free for now with a vague promise of riches in the future, we could use a 3d modeler. if you leave your email we could maybe work something out.

If you'd get the robot + a ten year service deal for $1M and it could do basic tasks equivalent to something close to minimum wage it will become affordable at $1M. Because it can work around the clock and does not require the same employer fees it will pay itself off withing ten years.

>>7107594
Cool, is your company doing well?

>>7107722
>I used to feel for losers like you who failed to become anything, but then I remember how big of faggots you are and how you deserve what you got.

Do I smell projecting? Where do you get that from? I'm in engineering, not being some comp sci fag that either is a code monkey or he does nothing for a job. Only a comp sci retard would think that's a good life, top kek

Why is /sci/ so mean? CS has many interesting subfields, such as artificial intelligence, information theory, computational physics, theory of computation... Why does /sci/ hate computer scientists so much? Any competent CS curriculum is 80% maths anyway, so if you complain about CS you might as well complain about applied maths in general.

>>7110235
inferiority complex

>>7109238
sure thing man. i'm pretty good with solidworks/catia.
dat.sonder@gmail.com

>>7109238
>>I am proud to say that we have solved that problem.
simulation or it didn't happen.

>>we have a solution for this too.
so do you have estimates of the mean time between failures or any calculations on how long these things will last?

>>7110235
Math Major Comp Sci minor here.

>Why does /sci/ hate computer scientists so much?

I think it's precisely because the field has such a high ceiling and a 50% drop out rate it leaves people a little bitter. I've notice some people just do not take well to code, and those are usually the hardcore math/science guys. That is not to say that programmers cant be hardcore mathematicians, many obviously are. I'm merely suggesting there is probably part of the brain in math that does not get worked the same way it would attempting to solve a problem in a high level language. There's a lot of creativity involved and to be a god tier programmer you need a god tier short term memory. Advanced math uses more long term memory so maybe that's it.

>>7110941
> usually the hardcore math/science guys.
I don't believe you.

The reason why I find it so easy to code is the same reason I find it so easy to do math.

>>7110965
>I don't believe you

Could be dunning-krueger. Maybe you've never worked on a project with tens of thousands of lines of code OR could just be that you have great logical faculties. That doesn't speak to how many concepts you can hold in your head at once or many of the variables that are different from math to programming. I think STEM people all like to walk around with the delusion that they would take easy to programming when in reality only 1 or 2 people per graduating class tends to be a natural.