Building a Flying Robot

[KaweS]

hi,
i'm new to robotics as well. my advantages are that i can program in C [for computers, but someone already did mention it will take me a day to learn in] and i can spend a bit more ;p
my disadvantage is that i didn't even build a $50 robot ;p
i built just a 'vehicle' based on 2 CDs, polyester body and a DC motor form my ex-CD-ROM ;] it can go forward and even backward if i change connections ;ppp
i know that's nothing, but i'm a bit ambitious and i learn really fast if i want to learn something

i'd like to build a robot which can fly [preferably with wings - not rotors] [see:

http://www.youtube.com/watch?v=U5hGYWVdzVk

] but definitely not r/c but autonomous. i want it to fly around me and avoid obstacles.
i guess i'll need lots of sensors and quite good ic.

ok - now the questions ;p
how can i tell the robot where am i? satnav or what? ;p i don't expect to come to me hundreds of miles, but let's say from my room to kitchen ;p i want to put for example plan of my flat and use wavefront algorithm - is that a good idea?
what kind of ICs could i need? i think lot's of memory and fast processing time - maybe 8 bit is not enough?
what materials can i use? what for body/wings?
what about flying time? i don't want it to run just for 5 minutes... i know - more power - heavier batteries
what about costs... i know i cannot expect too good news regarding this point ;p

to be honest i'm thinking already about version 2, but for now that's enough ;p

thanks a lot for any help and ideas ;]

cheers,
Karol

[skatj]

Umm, an ornithopter, especially of that size, and indoors, would probably require a great deal of experience before someone would even think about working on it.

I suggest starting with a fixed-wing aircraft based on an inherently stable airframe that uses a GPS receiver as sensory input.

expect to spend 600$+ for that.

[frank26080115]

Although your enthusiasm is great, you need to crawl before you walk and walk before you run, and a flying robot is like riding a bike if running is compared to a home built hexpod.

If you must make a flyer, I'd suggest making a lighter-than-air robot like a blimp, exactly the same as a wheeled robot except way less mechanics to worry about, just keep it light.

[SmAsH]

a blimp? as in filled with helium and the motors underneath? thad't be cool

[KaweS]

i expected those replies. i know that is a huge jump over all the basics which i should practice, but i think i'm able to do it. but on the other hand i'm not in rush so i can do some more basic robots. just suggest me what could be helpful - i don't want to build a car with basic sensor just to learn how to avoid obstacles. do you really think it is necessary?
what more - i don't want it light - it has to be quite heavy, cause i want it to fly outdoor... so if that is possible [i could require really big wings] i want to fly it outdoor. if not i'd rather go for rotors. i want to keep it as small and heavy as possible ;ppp i know that i'm asking for impossible... and don't forget that i want it to fly for quite long time [1st post]. so if it just can't be done - i'll stay with light flier for indoor use only.
what about a blimp? nice... but not for me. i want it fast in movement and changing directions and i think it's impossible with a blimp.
i live and work in UK, so 600$ is not a big deal for me ;p

cheers,
Karol

[SmAsH]

lol yer you brits can afford everything of ours. anyway...if you want it heavy i would suggest atleast 4 powerful motors with BIG blades. you may be lucky.

[airman00]

quadricopter or tricopter

first worry to make it levitate autonomously , no moving yet

have the blades tilt for propulsion

[JonHylands]

If you're set on using wings for flight, I would strongly advise you to buy one first, so you can get an idea of how well they work, how much they weigh, and how the wing mechanism works.

http://www.hobbytron.com/CyberHawk-Ornithopter.html

Whatever route you take, to get to where you described in the first post, expect to spend at least a couple grand and a couple years...

- Jon

[gmatkins]

You might be able to hack a WoWee dragonfly. They're cheapish and they sure fly. To save a little weight try putting it's brain somewhere off-board and have the brain talk to the body via RC or IR. If you were willing to wear they transmitter on your body (and it could be quite small) then sensors on the dragonfly's body could use the signal both to receive information and to keep track of you.

[HDL_CinC_Dragon]

If im not mistaken, you said you wanted to make something that flew like a regular airplane as opposed to a helicopter? That wouldnt work in and indoor setting. It would crash into everything and anything.

Basically for autonomous navigation and avoidance like that, you would probably need image recognition rather than just range finders. Range finders can be fooled and mislead. That is very very advanced. Like some of the others said, first work on making something that can simply rise straight up and then land again by going straight down. Perhaps using a range finder on the bottom of it so it knows when its about to hit the bottom so it can ease its velocity to about .1 m/s or so to reduce the shock of landing... and also some compression springs on the landing gear to help with that also.

[KaweS]

thanks for all the responses ;]
just a quick note to be clear what i'm looking for. i want a craft what can hover and move all the directions like a helicopter, but i'd like it to use moving wings, like a bird.
cyberhawk looks cool and i love one point from it's description - 'So realistic, you'd swear it could drop a load on your head!' hehehe
anyway i think it will be to difficult to develop and take too much time on experiments on most efficient design, so i think i'll stick to rotors.

1st problem which i need to solve is the design. do you know any proven architectures? [yes - helicopters] ;]
2. batteries - i read a bit and i found out that the LiPo batteries are the best for air-crafts. please correct me if i'm wrong.

the idea with brain out of the body is quite good, but where could i mount it? carry it with me? what about transmission? wi-fi? and i guess i have to put some basic intelligence in it in case of lost connection - i cannot allow it to crash ;p
what about GPS? i read a bit about it and hi-end receivers can pinpoint object to 2-5 meters, what i think is not enough. for instance if i'm in another room and i want it to come to me it will not know where to go... what about levels? what if i'm above or below it - on another floor? i live in a maisonette...
any other ideas for positioning?

//EDIT
[youtube]sN7jz7P7p1Y[/youtube]
i just found this vid and i think that's just awesome. i didn't know that something like that is even possible. how can it hover? shouldn't it just spin around? that's why helicopters have tail propellers...
that would be great for my robot. it would move vertically when precision is important and horizontally to go fast

[dunk]

hi KaweS,
your not making things easy for yourself.
what you are describing is taking 2 of the harder unsolved aspects of robotics and combining them.

indoor localisation is not an entirely solved problem. on a stable ground based platform it takes people years of research.

autonomous aircraft are very difficult as well. people have managed to make rotor platforms that can keep them selves level but it a non trivial task and flight times are very short due to battery life.
again, be prepared to spend a lot of time and effort building a platform that can take off and keep it's self level.
look into balancing segway style robots for more information on the types of sensors you will need to use.

how can i tell the robot where am i? satnav or what? ;p i don't expect to come to me hundreds of miles, but let's say from my room to kitchen ;p i want to put for example plan of my flat and use wavefront algorithm - is that a good idea?

what about GPS? i read a bit about it and hi-end receivers can pinpoint object to 2-5 meters, what i think is not enough.

2-5 meters is only out doors. satnav is no good indoors. if you can find a sensitive enough GPS you might be able to get resolution down to 20-25 meters or so indoors.
if you are prepared to drop your indoor requirement then GPS may be an option for you.
there are no easy ways to determine your robot's position inside.
the easiest method would be to fill your house with beacons and have an onboard map of your house and where the beacons are.
another method would be to use ranging sensors and use pattern matching against an onboard map of your house in an attempt to determine where your bot is. this approach is hugely processor intensive and difficult to achieve, especially with the limited lightweight sensors you will need to use on an aerial platform.
bear in mind you are going to have huge problems getting accurate sensor readings from a moving flying platform.

although i'm not suggesting you use it, search this forum for SLAM (Simultaneous Location And Mapping) for a little insight into some of the problems you will face.
you will face similar problems with the advantage of already knowing the layout of your house in advance so you can start with a known map.

.what materials can i use? what for body/wings?
what about flying time? i don't want it to run just for 5 minutes... i know - more power - heavier batteries
what about costs... i know i cannot expect too good news regarding this point.

look around on Radio Control forums.
top end ready built electric RC helicopters can run for a maximum of around 30 mins on LiPoly batteries. i would expect far less from a home made platform.
as for price, this is going to cost you many $thousand as there is no way you will get it right on your first attempt and will undoubtedly end up re building many times.

as you might have guessed, i don't think you will get very far treating this as your first project.
i think you would be far better working on the different challenges independently with the aim of building a flying bot once you have some experience in the technology involved.
as a first project start trying to build a working microcontroller platform that can control servos and RC motor controllers. for someone with the experience you describe this will be non trivial.

next use it to control your wheeled vehicle so you can experiment with sensors and map solving on a ground based platform. i'd be surprised if you get a ground based system that can locate you on it's own within a year.

next i would experiment with angular rate sensors and acceleration sensors to try and build a balancing segway style bot. you will need good understanding of these sensors to keep any flying platform upright.

anyway, good luck with it but seriously, you will have far more success if you pick simpler objectives to learn about the technology involved before starting on your flying bot.


dunk.

[KaweS]

i didn't realized it could be that difficult...

do you think that i should build a heli first and then try to create a plane from it? [ok - i'll create a wheeled vehicle first, just to learn basics, but after that]

indoor localisation.
i'm thinking about making my robot able to recognise area - the room where it is at the present moment. i have no clue how long would it take and how much memory consume... [i think about... let's say 3-5 inches cubes] it should remember some areas even after switching it off [all rooms in my flat, my workplace...]
when it has a 3d map of area and it's able to go any direction i think it will not be to hard to find and go the quickest way from A to B.
i think i need to use different sensors altogether to create such a map. and it has to move around to recognise all the area. it needs to be able to check the way it goes to avoid hitting moving obstacles. what about tracking other objects? if it detects collision course it needs to find another way or just hold. but this would be very difficult and consumes lots of power...
i can install another thingy on me to recognise the area and then send info to the robot which obstacle is me ;]

what kind of sensor would be useful/necessary?
accelerometer for sure
color sensor? i think it's useful, but not necessary to work
compass? i have no clue what could i use it for?
encoders - yes
force and torque sensors - yes
Sharp IR Range Finder - lots of ;]
sonar - i think sharp ir is enough
something else?

[airman00]

first make a quadricopter or tricopter that can hover autnomously , then well talk 

[HDL_CinC_Dragon]

~To affectively achieve what you are talking about here, I think what you would need is cameras and image recognition... which is a huge deal.
~You would need an accelerometer of course, yes. To determine how fast it is going parallel AND perpendicular to the ground (2 seperate accelerometers).
~Color isnt really TOO important for a robot unless you want it to track and follow and/or shoot at a specific color... or perhaps I identify specific color coded markers.
~A compass would be required so the robot knows its heading in relation to the map in its memory
~The encoders can be used on the rotors (of a heli, please dont even try to make a plane out of it for indoor use. That could be disastrous) to monitor the RPMs. This is handy for a number of different reasons.
~To get the amount of force you will have to plug the amount of RPMs into an equation involving the amount of air moved per rotation of the blades and the amount of rotations per minute(or second).
~For the Sharp IR, this could be used to sense the distance between the bot and the ground, ceiling, so it doesnt rise or fall into anything... like a ceiling fan or something ...
~Sonar, IMO, would be better for use on the front and rear of the craft so that it can detect objects in front of and behind it so it doesnt try to move forward or backward when something is in its way.

[KaweS]

thanks for your replay - it helps a lot ;]

but do you really think that sonars will be better than sharp IRs? as i read IR is not perfect, but still works. in the opposite sonars are very vulnerable to noise, in particular in places with not too much space around and lots of obstacles.

what about accelerometers, they can be even 3-axis. look what have i found on google: http://www.rapidonline.com/productinfo.aspx?&tier1=Electronic+Components&tier2=Integrated+Circuits&tier3=Sensors&tier4=Digital+MEMS+3-axis+accelerometer&moduleno=79653&catRef=82-1006 and it's just £14. i just wonder if they work all the same. i know that i need hi-end cause it has to be quick and hi resolution response, to react fast.

compass? i think if it has a map, it is enough, because on the map doors will be located as well and depending on readings from sensors it will know where it is at the moment and which direction it has in front of it. compass could make it faster to recognize at the beginning, but it's more to implement, when after few readings from sharp IRs or sonars it will localize itself as well.

cheers,
Karol

[HDL_CinC_Dragon]

Dont have much time to explain at the moment because I have to leave but:
The sharp IRs are very much directional but the sonars have a wide angle of detection when in a stationary spot

I would have the compass still. I wouldnt rely on the positioning of objects to find your heading. This is extremely difficult not to mention very system intensive.

[SmAsH]

but would a scanning ir work?

[cooldog]

but would a scanning ir work?

yes but i's really heavy for a flying robot

[airman00]

well today i was playing around with my r/c heli with a camera mounted on and I realized that if you could have accelerometers relay info back to a PC over bluetooth or something , then the robot can autocorrect for human errors. If the heli sees its going to crash it can autocorrect the human operators mistake.


Or even easier, since I have a wireless cam on it , I can output the values of the sensors onto an LCD on the heli and then use roborealm to convert the visual numbers into data numbers.
http://www.roborealm.com/tutorial/Digital_Reader/slide010.php

What do you think?

[KaweS]

i need to say that i'm a bit surprised of myself. to be honest most of my projects [any kind of] i never finish when it comes to learn new stuff ;p but today i spent many hours studying helicopters - how do they work and so on. generally last week i learnt lots about robotics and related stuff and i still want to learn more ;p i just printed many pages regarding helicopters and hopefully i'll have enough time to read it at my work ;p

i think that i'll buy an RC heli to hack it or at least some parts of it - rotors and engine
i think i'll skip a quadracopter or a tricopter. tomorrow i'll read more about various types of rotor systems and then i'll decide.

i still don't know what about sonars and IRs - instaling 8 around plus another two - one on the top and other at the bottom or some sonars - no idea how to locate them.

compass - honestly i don't really think about it now.

camera on my heli? for sure, but not yet ;]
airman: keep developing and share your results so i'll use them when it will be the proper time ;p

[airman00]

well can I just get four r/c helis and connect them together and use those main blades?


something cool

[airman00]

forget four blades


would two blades work?

basically two helicopters attached together

[KaweS]

i don't know too much about rotor systems yet, but
2 main rotors connected together - no - they both spin clockwise so you cannot use it.
2 helicopters connected together - somehow it could work - try to connect two the same helis, which work at the same frequencies, to each other and use one transmitter to control them like one - could be really nice ;] i'll try it when i'll be able - my manager won't be around and it won't be too busy

lil'bro looks crappy, but has nice abilities ;] try to find the frequency it works at and take it over hehehehe

[HDL_CinC_Dragon]

lil'bro looks crappy, but has nice abilities ;] try to find the frequency it works at and take it over hehehehe

That would be hilarious

2 main rotors connected together - no - they both spin clockwise so you cannot use it.

You could reverse the polarity of one of the choppers' motors and flip the prop of that same chopper upside down. That would work.


That would basically create an Osprey

[airman00]

no i mean a two bladed chopper like this


top view would look like this
O-O

the O being the blades

but try the two helis attached , tell us what happens

[SmAsH]

hmm yes that may work but youll probibly have a hard time changing the frequency.

[hgordon]

If you're serious about this, spend a few hours (or days) reading old posts here -
    http://www.rcgroups.com/multi-rotor-helis-200/

The problems you want to address are VERY complicated, and you need to develop a basic understanding of flight dynamics for different airframe architectures before you can begin to figure out how to integrate sensors and onboard computer controls.

That said, I will note that this is a VERY interesting subject area, and quite addictive.  I have 3 UAV projects in the works at the moment - 2 of them are described here -

    http://diydrones.com/profiles/blog/show?id=705844%3ABlogPost%3A23537
 
    http://diydrones.com/profiles/blog/show?id=705844%3ABlogPost%3A22738

As you can see, I'm still struggling with basic flight stability issues, so fully autonomous flight is still a distant goal, but I'm motivated.

[HDL_CinC_Dragon]

no i mean a two bladed chopper like this
top view would look like this
O-O

So it would create a Chinook

[KaweS]

i've got some cool stuff for you guys
just gimme few minutes to upload it ;ppp

ok
here it comes 2 rc helis connected to each other and operated by one transmitter ;]
[youtube]KWnobua4cGQ[/youtube]

more info: http://robo.kawes.eu/picoo/
ps: this time i compressed the imgs ;p