Year 12 Systems Engineering Bi-ped project

Hey all,
I thought i'd share my year 12 project that's currently in the makings. It's a Bi-ped, I'm planning to have obstacle avoidance on it as well, the actuators shown in the sketch up plans are powered by air but I'm planning on designing my own actuators the run of a linear track system powered by low rpm motors. I'm having some trouble on deciding how the best way to to do that is, and of what control system I should use, so far I've got it down to either picaxe or ardunio. Any advice/guidance on what to do is very much appreciated

Basically I need help with:
- How to build the linear actuator's
- Picaxe or arduino (Or something else I havent considered enlighten me )
- A good cheap on/off switch box for the power and 6 motors, so with 7-8 inputs/outputs
- General comments on the design





Any input is helpful
Thanks all

 

Cool project, but when you say you're going to build the biped, design the actuators and are going to implement object avoidance (of course which is no use if the biped can't maintain self stability, in itself a problem already), the first thing that pops into my head is 'bitten off more than you can chew'.

But that aside, I'll give you some pointers about the way I see it:

Linear actuators:
The ones in your CAD drawing are your typical industrial style 'electric cylinder' - essentially an acme leadscrew or ballscrew that gets forced in and out by way a normal screw works. You can reduce the complexity and cost from this design by using allthread and a nut and gears, but the smoothness, loading performance and other factors all suffer somewhat. There is however an small out on this path, there are commonly bottle style jacks which are hand wound (sometimes electronically wound) that would be quite suited to adaptation - but heavy, quite slow and far from inexpensive.

Another option is a rack and pinion system. There are sets sold on element14 and RS components which are designed to work together, and will probably work for these purposes.

Actually building actuators from scratch is quite a project even with premade components such as lead-screws and general purpose gears. There are a lot of in depth design details that you'll probably miss that you would have never realised were 'a thing' unless you had mentor that specialised in those particular styles of equipment. Also consider your limitations in manufacturing techniques. For example do you have the right tooling to make something that needs to be concentric actually concentric?

Therefore I recommend you reconsider 'building' the actuator, and go with something pre-made for this part of the project. There will be a variety around at different price points, particularly pneumatic gear from taiwan or china will be at quite accessible price points and should do what you're looking for. Other places to look would be enthusiast robotics places like pololu and sparkfun, as well as hobby places like hobbyking and r2hobbies.

microcontroller (i.e arduino/picaxe)
If you're serious about 'obstacle avoidance' as in humanoid robot obstacle avoidance, you're seriously looking at kinect+PC type hardware. But I'll give you the forewarning that this type of environmental navigation work is really at the undergraduate university level type stuff and even at that extends further. This might (will?) not be feasible for your year 12 project.

On the other hand, if you're just looking at getting this thing to be kind of stable, then that's mildly easier. Arduino's probably got the power, and is a generally well supported platform, and will be technically capable of getting the job done interfaced with one or two gyroscope sensors and maybe ankle force sensors for more fanciful work. But here we're assuming that you have a good grip on control theory as well as the mechanical dynamics of the system you're intending the build.

A good cheap on/off switch box for the power and 6 motors, so with 7-8 inputs/outputs
IMO only a pretty small problem from the whole scheme of things you've chosen, but you're looking at putting some relays in a project box, or for the motors perhaps motor drivers (check out the kits at the aformentioned sparkfun and pololu)

General comments on the design
My concern regarding the CAD drawing relates to the lack of lateral pivot of the legs or ankle. If they were spread out then that would be much easier to balance. I have a suspicion that it will be somewhat unstable on that plane.

In this sense, as well as the notes above a triped would be a much more realistic project to embark on, although less cool.

I worry when you mention cheap, because it's probably not going to be.

Accurate CAD is probably more important than you think it's going to be, because inevitably that 'thing that doesn't look right on the CAD won't be in real life' will come back to bite you later.

Anyway, hopefully that's given you some stuff to consider. I don't know what's normally expected of the product of your projects (perhaps this is a group project I don't know?) and your specific experience and capabilities in manufacturing and design, but you'll need to assess what's the best route to choose for you. Well hey you're learning about systems engineering so this is just part an parcel of it

 

I agree, this does sound more like a university level project.
Bi-pedal movement is quite difficult unless it just shuffles along. It needs lateral movement in order to tilt side to side to balance (then you'll need control loops such as PIDs running on the MCU in order to correct the movement, taking error into account, etc - which you'll learn about in early university mechatronics courses).

Like mtma said, building something that does not require any balance control while moving would be much better. Also consider a robot that has wheels.

Taking balance out of the equation allows you to focus more on the obstacle avoidance and other navigation.

Obstacle avoidance: For basic functionality, consider ultrasonic sensors - one to transmit, one to receive. i assume the closer something is, the higher the input will be to the receiver (ultrasonic waves reflecting off the object surface).
- this doesn't work overly well with soft furry objects by the way (e.g. cats) as they absorb a lot of the waves.

Also consider light detection and line following.

Arduino gets my vote too.

I also know of schools that use Lego Mindstorms kits to build soccer playing robots.

 

this does seem very ambitious, My engineering degree is in robotics/mechatronics and this would be ambitious for a final year university project - for cost reasons if nothing else.

the design in your OP will only work if you have a swinging weight on the top of the platform as the legs only move through one plane and have no side tilt etc, it could work if you made 3 pairs of legs but turning would be pretty inelegant.

go with arduino for sure.

 

How about using servos instead of the expensive actuators you can get large units these days. Anyway have a look here http://www.mcselec.com/index2.php?option=com_forum&Itemid=59& there are a lot of very good projects here. Also a lot of very helpfull and experienced people a lot of the code already here would be helpfull to you eg code for triple axis gyros and auto pilots

 

I agree with this.

RC servos are cheap, trivially easy to drive (just give them 5V and a 50Hz PWM position input), and designed to make mounting easy. If all else fails, you can just hot-glue them in place.

Larger (ie non-RC) servos are available, but they do tend to be rather expensive.



Have you thought about doing a hexapod or something similar? They're much easier to balance, and there are already lots of designs using RC servos. For certain geometries (which happen to make joint solutions easy to compute), software is available to drive them. Designing and building one suitable for a specific task would be a fine Year 12 project.

Edit: there's a reason that 99% of robots don't use legs. Honda's Asimo is probably the best-known bipedal walking robot, but it's got an incredibly complicated walking system (sensors all over the place, 12 DOF actuation in the legs alone, predictive control, etc). The best-known walking robot (removing the bipedal restriction) is probably BigDog, and that's similarly complex (4 DOF actuation in each leg, more than 50 sensors). If you want simple control then wheels are the way to go.

 

Yep, this is the bit that grabs me - unless you design something which is inherently mechanically stable (Hint: Most bipeds aren't), you're looking at a non-trivial control theory problem - something quite akin to the inverted pendulum problem that most students of control theory encounter pretty early in their studies.

Heck, I'm a fourth-year electrical engineering student, starting to specialise in Embedded Systems and Control, and this thing intimidates ME - frankly, I suspect the only reason it doesn't intimidate the OP is that they don't have a clue how much is involved.

One other thing: The design as shown will never work, the mechanics make no sense at all.

 

It is quite a hard project, but the clause of the project is that if it doesn't work, i just need to explain why and still get full marks ;D That aside, I still wish to get it as close as possible to working.
After reading your post i think i might go with pre-made actuators but i would rather a rack and pinion electric motor driven one than gas/air driven pneumatic, any ideas of where to source one?
kinetic would be too hard to integrate (and most probably out of my skill range) the robot it self will be more of a simple shuffle movement than a total flexible joints for stability, my idea (not shown in the cad designs) is the have a movable ballast on the top, would that be feasible in the overall design?

Thanks heaps for the reply mate, was really useful

Yeh I get what you mean, I am going more for the shuffling along type thing with a movable ballast on top for stability (If feasible) and i plan to embark on this project again in my uni project with full flexibly and movement
Ultrasonic sensors sound like a good way to go thanks for that
Is line detection hard to program in?
Looks like Arduino is winning

Thanks for that cheers mate

Arduino wins for sure now :Thumbup:
Yeh if I'm only having the movement in one plane, a movable ballast should keep it balanced correct?
Cheers

I had a look at servos, Would ones powerful enough be able to move the bi ped? Its about 85cm tall, its not a small thing, All out of aluminum with big 5kg battery's on it + ballast, I didn't think servos would be able to cope with the weight, I'm in the process of looking through that forum thanks for that I think it will come in very handy

Yeh i see your points it will be a hard project with two legs, but I think in one plane without the flexible joints and with a movable ballast it should (hopefully ) be achievable, As i said above the bi ped is fairly weighty and i don't think the servos would keep up and the actuators would probably end up being cheaper
Cheers mate

Your right i dont have much of an idea of the control theory, hence why I'm putting it in one plane with a movable ballast as the balance, seems like a medieval way of doing it granted, but as you stated theres no way i have the skill/time frame for something of that magnitude, I think the designs i have up there (albeit poorly drawn) should work for a one plane bi ped
Thanks mate

 

What's the budget?

You can get linear actuators on eBay, but they're not cheap (about $50 each plus $50 shipping). On top of that you'll need a suitable power/control circuit and feedback loop, since none is included.

If the budget is less than about $1000, those are going to eat the whole thing straight away. In that case, I'd use RC servos and design the whole thing to be much smaller/lighter. Easier to transport, easier to repair, cheaper to build, etc.



Regarding line detection - are you doing it with photodiodes/phototransistors, or are you using cameras? The Arduino can handle photodiodes and phototransistors easily, and line following with those is straightforward (on a wheeled vehicle, anyway. Bipedal will be harder). If you're doing it with cameras, then what you need is a Hough transform and a lot of processing power (ie a PC, not an Arduino).


Regarding marks - when I did a similar course, you could certainly get good marks without completing the project. However, you needed to demonstrate that you had done a fair bit of work on it. This is where I think the problem will lie with yours - you can do a huge amount of work and have virtually nothing to show for it, because the project itself is so large. Getting to the end of the year and saying "well, I was going to build a bipedal robot, but what I actually did was get an Arduino to move an off-the-shelf linear actuator" does not look good. However, I wouldn't be at all surprised if that's as much as you've got time for in the year. If you haven't done any programming before then even getting to know the Arduino will be somewhat tricky. Then there's a whole lot of circuit design to power the actuator, getting PCBs made, soldering everything on, repeating at least once to fix the inevitable bugs in the first version, and so on.

It'd be much, much better to say "I'm going to build a closed-loop controller for a small actuator" (note that separating this from the robot project allows an actuator that can be connected to the Arduino much more easily than the big 24V one can) or "I'm going to make a simple wheeled robot follow lines and avoid obstacles". Then (a) you don't get stuck trying to do ten things at once, with the result that you do none of them, and (b) you'll hopefully be able to finish the project reasonably early, leaving plenty of time for writing it up and doing work for whatever other classes you've got.


Edit: what's the foot actuator actually for? Wouldn't it be easier (and cheaper) to mechanically ensure that the feet remain flat? Then you'd reduce it to 2 DOF per leg, which is what you need (ie up/down and forward/back movement).

Have you thought about how it's going to do corners? That'll mean having both feet on the ground at once and dragging itself around. It'll become very unpredictable because you're basically relying on a loss of grip without any way to control where/when that occurs.

 

I can't agree more with this... I'm currently a final year mechanical engineering student and this is probably what you'd do in a group of 4 for a final year project, and thats with decent knowledge of automatic control, years of fundamental mechancal design, and still it would be no easy feat.

 

Im a 4th year mech eng student and completely agree. I suggest make up a little model using maybe aluminium sheeting and string and springs for the actuators. Pulling on the string or compressing the springs will let you know if the design will move like you want it to. It will also tell you what kind of logic needs to go into the control system. ie this one needs compression etc.

I look at your design and it kind of makes sense but there are a lot of factors that need to go into your control system and that could take most of the time. The first issue I see is getting the actuators to move the joints. They are on such tight angles with the legs that they will barely move them but instead they will just try and compress or extend the leg material. I would build the model and play with the angles to get it right. You want the actuators to move the legs not compress them. I hope this makes sense and helps. Let me know if it doesnt.

Good luck.

 

I've been looking at these for some time, they may give you some ideas for your project
http://www.youtube.com/watch?v=cuEuc4hiy3E

there are also larger soccer robots as well, all bipeds

http://www.youtube.com/watch?v=wpNaANCq0NM

 

Without commenting on anything else... if you're down to a choice between Arduino and PICAXE you definitely do want Arduino.

 

I have nothing to add other than good on you for giving this a go. When I was in year 12 I just got my amateur radio licence and I thought I was pretty clever for doing so.

If I was trying to do something like this when I was 16 I would have curled up in the foetal position and started to cry!

 

Have a look here at http://www.pololu.com/ they have all the parts, but you will need a budget of about $3000 to $4000 for such a project. Whilst in year12 (I am over 50 now) in my industrial arts group we did a full structual analysis of one span of a bridge in sydney called the Iron Cove Bridge we got the plans from the then RTA and found it would fail, after getting a new set of plans sent to us from the RTA with additions to the truss and photos of a failed span! we had confirmed our calculations and nice letter from the Cheif Engineer. So Humpty I think its great that you want to try but look at your budget first. Note the site I sent you to the programs will run on Arduino hardware but its less expensive to build your own I know this from experience. Also with your design as is will only be able to lift its leg have a look at nature, your own leg! it needs to move forward and swivel , have a look at a backhoe its joints and its hydraulic rams and so on. There is a four legged backhoe that has legs not wheels to walk over fencecs in yards its legs are small versions of its digging arm.
For inspiration have a look at this http://www.youtube.com/watch?v=UjxEj8_A0Yk

 

From what I had gathered in the completion of my systems engineering project the harder projects seemed to not get completed in the restrictive time frame of VCE. Having said this though I chose to design, construct and debug a hand made poly carbonate controlled tank with pan tilt camera and wireless pc control with the use of an arduino and an xbee. I guess the thing is to push yourself but be conscious of what you think you can achieve. Aim for the stars but settle for the moon sort of thing. Too many people aim for something that looks impressive and achievable but in the end they forget to factor in time lost while waiting for parts and what not. The last thing you want in VCE is to have a half completed project with only weeks to go (I have been there. Somehow I made it and received full marks ) and with NO amount of reasoning could you have received full marks if to not complete (From what I remember, check on the vcaa).

As far as microcontrollers go if you are able to follow a basic form of C++ or are willing to give it a go the Ardunio would be more than capable of doing what you wish and with the aid of new programs for visual programming (flowchart/block style) it is defiantly achievable. But if you prefer a much simpler option the picaxe would do just fine too. My vote would be for the Arduino.

For a control system I would look into designing and fabricating your own motor driver board to work alongside your chosen microcontroller, or at least consider it in your report.

P.S. Keep an UPDATED work journal containing EVERY SINGLE problem that occurs along the build it will make the report SOO much easier to write, trust me.

 

Budget is around $300-400 maybe $500, yeh linear actuators are expensive, hence my attempt to make my own,
I'm not sure how I'm going to do the line direction system yet, I'll basically do it with whatever arduino can handle, so basically i'll be using photodiodes and phototransistors,

Yeh with the marks, i'll be taking pictures of every single thing I do, programming and all, if it doesn't work i'll still get close to full marks, I've already discussed that with my teacher

I haven't considered the turning yet, but as I stated it will be single plane bi ped, basically a mechanical loop making to go forward in a biopedal fashion, it will be reliant on loss of grip, but I plan for to only be used on flat surfaces and I think the loss of grip way should be fine.

Cheers man

It's a single plane movement, no control theory/automatic control theory is really required, It may be ambitious but I think it close to within the limits of what I can do.

Thanks mate

A model, thats a good idea, will do that thanks man, I get what you mean about the actuators not compressing the legs, i'll put a bit more research into the angles on which they are mounted.

Thanks heaps

They were very informative and fun to watch cheers man

Safe to say Arduino wins now aye ;D cheers

Thanks man much appreciated, but I'm sure ill be crying in the fetal position before to long...... haha cheers

That video... I may have watched 5 times in a row... that's awesome. Thanks for the link ill be sure to sus it out Wow the poor Cheif engineer, showed up by some year 12's nice work man At the moment im not planning to make it quite that complicated, just a single plane, which alleviates most of the swivel movements and such.

Thanks heaps man, much appreciated

Well, the moons not all bad there's rocks to play with.. Well I wont be expecting 100% if I don't get it completed put it that way, but according to the vcaa, as long as I can fully explain why it failed/didn't work I can get 95% of my marks,
I've done basic C++, but I'm fairly terrible at it, there's of pre-done programs I can modify from forums some guys have suggested but I'm more than willing to pull a few all nighters to get it programmed, I reckon that side of things will be fine

I will consider it, but with the time constrains I don't really think its feasible as much as I want to do it :/

Yeh I end up writing a page or so every lesson

Thanks heaps man




-Thanks everyone for the input, its been very informative and very helpful, cheers lads

 

I would suggest making it smaller, about table top sized. The costs will be significantly lower with components that big and could handle the loadings required. (e.g servos)

You might want to check out the Robotis Bioloid robot model for some inspiration, but you don't neccesarily have to go with the way they did it, you can still use the tendon style actuation at a different scale.

 

i'd look into a chipkit - which is a pic32 arduino clone compatible with arduino shields and has a modified version of the arduino software environment available for it.

Documenting - the devil is in the details, i tend to live notes and comments in shorthand before i attack the diary.

 

Sounds good.

It might be worth clarifying that.

If the robot is built and almost working, just missing one or two things (eg. last-minute code problems) then that's certainly going to get a lot of marks. That would be an extremely impressive project.

If the robot, at the end of the project, consists of a single home-built linear actuator that can be connected to an Arduino, then I can't see how that would get many marks. However, I can definitely see how that might be as far as the project gets in a school year. Getting that much working means:

• You've figured out the Arduino coding environment
• You've either built a DAC for the Arduino or sorted out PWM motor control
• You've designed, built, and populated the PCBs for the motor drivers
• You've understood the basics of control theory and tuned a PID controller for the linear actuator.
• You've selected a suitable motor for the actuator, taking into account starting/maximum torque, voltage, current, etc.
• You've located some suitable lead screws or threaded rod, and figured out how to attach those to the motor
• You've designed and built the housing for the actuator.

All of the above will most likely need to be repeated at least once, possibly 3 - 4 times in order to get a working actuator.

Okay. Might be worth testing that at the first opportunity. My intuition is that it'll be really jerky and unpredictable, which will be a pain in the neck for line following.

I think you may be in for a surprise.

The actuators alone will require a fair bit of control theory. All they provide is a torque input (motor current) and a position output (potentiometer). You need to add a controller that takes a desired position input and a measured position output, then continually computes torque inputs. Using hobby servos would avoid this (they have a built-in controller) but they're not big enough for your target size.

I don't think that you'll be able to get away with a simple "turn motor on for this long" system; the forces will be changing too much for that to work with any sort of reliability.