View Full Version : [Contest Entry] AluHex: 4DOF Hexapod on a Budget

10-19-2011, 11:10 PM
Hey everyone, this is my build log for a project I will be working on, called AluHex (unless a better name comes along, I'm not particularly attached).

I've actually never been a huge fan of hexapod robots because they tend to be extremely expensive and when it comes down to it, they tend not to do much other than look cool. That being said, looking cool is pretty awesome in my book and I hope to build a hexapod that not only looks awesome but that also gets by on as little gear as possible.

I'm also hoping to show off what can be done with DIY tools and open source hardware. With all the extremely cheap import products available nowadays it is possible to build some really nice machines extremely cheaply. This will actually be my first from-scratch robot build so there will be a lot of learning along the way.

The biggest cost savings will be by skimping on the servos. I will be using twelve each of the following rc servos:
HXT900 9g / 1.6kg / 2.69$

HK15138 38g / 4.3kg / 2.83$

In total, the servos are 66.24$ + shipping. I'll call it 100$ flat. This choice is probably the biggest risk I am taking in my build because there will be a lot of time invested in the project that will go to waste if servos start burning. I hope that a combination of light weight and smooth gait will allow the hexapod to walk smoothly.

While these servos are very cheap they are reasonably well trusted among the rc community and if a single servo fails on a plane you are in trouble. I'll be able to comment more on their suitability as the build progresses. The servos will be powered by a pair of 1Ah 2s LiPos regulated down to 5 or 6 volts.

Choosing how to control the hexapod was a somewhat hard decision but I believe I have made a very compelling choice with a LeafLabs Maple Mini + Pololu Servo Controller:

(http://www.pololu.com/catalog/product/1356)One of the main criteria I had when choosing a micro-controller was that it needed to be 32 bits and small. The pololu controller is a bit pricey, especially considering the maple mini alone could control 24 servos but I made this choice because it will greatly simply both programming and testing. I am no entirely confident in my ability to program whatever interrupts would be necessary to control 24 servos while doing the IK math and the pololu board will also help with the wiring and testing via computer.

The combined 85$ compares favourably to all the alternative solutions I have seen. There will also be an XBee thrown into the mix for some kind of remote control.

I hope to keep the entire project under 400$ and I'll go into some of the mechanical plans later.

Cost to date (Adjusted for shipping):
100$ (Servos) + 125$ (Electronics) + 30$ (Batteries/Regulator) = 255$

10-21-2011, 03:21 PM
Since I'm using hobby servos of varying sizes there aren't really any brackets I can use (plus I'm not necessarily willing to pay for brackets). With 4DOF the robot will have some weight so it is very important that the joints are supported on more than one end. If the joints are only connected by the servo horn they won't be very rigid.

I've made some basic plans on paper and the final robot will hopefully end up looking something like this:

Most of the segments will be made of three pieces using the servo horn and a bearing on the opposite side.

In terms of material, the hexapod will be made of primarily aluminium with some acrylic and carbon fibre thrown in for looks. My three favourite materials.

Only a small amount of this aluminium is for the hexapod. I will be using mostly 1/4"x3" flat bar as well as 2mm acrylic.

Also a nice stack of bearings and M2 socket head screws. Most of this stuff is from HK. The sheet of carbon fiber was 10$ and the bearings are cents.

Total Cost:
255$ + 70$ (Material) = 325$

10-21-2011, 04:57 PM
Cool project. I'm a big fan of them myself and have built a few in the past.

I switched out the HS-55 for the HXT's. Same pattern but a little beefier.

These hxt900 servo's are actually a great buy. And the servo horn is quite strong.
You could save your self some weight by not having to box every joint and going with aluminum around 1/16 thick.

10-22-2011, 04:43 PM
Thanks, your wasp looks really cool and its good to know the servos should be able to hold up. If I need to I'll switch out the servos for hs-55's. I've also noticed that HK has released a digital version of the hxt900 in both 5 and 6V versions which have more torque and ball bearings for about 3$.

I'm hoping to make a hexapod that looks really "solid" so I'll stick to boxing most of the joints for now.

The vast majority of the parts are going to be cut by a mill/drill I converted to CNC. I'll be starting at the end of the legs and moving inwards, doing the CAD/CAM as I go.

Here is a pair of legs in the cam program:

And then the g-code ready to cut:

Get some material cut up:


One side machined.

Second side getting machined.

Finished parts.

I'm hoping to keep the visual style consistent for the rest of the parts. The mill gives them a pretty snazzy finish that is shiny in a really cool way.

Although the mill does a lot of the work I still have to do manual toolchanges and the cycle times are fairly hefty. I'm planning on making one leg first to see how things look before doing the full set.

10-22-2011, 07:11 PM
Don't switch to the HS-55's. They are more accurate (they don't glitch) but they were made for light weight airplanes/heli's. They have weak gearing inside that I constantly striped. I have yet to strip a gear or wreck a internal circuit board on a 900. As for the digital version I didn't know they made them but I'm going to give a few of them a try now.

Nice machining. Legs look awesome.

I see that you use Mach3, I use the same program. What cam program is that?
I noticed the mill looks like some type of model from Grizzly? That green paint looks the same as my G0463 which I tore apart and heavily modified.

10-22-2011, 10:53 PM
The mill is an RF31 clone which is equivalent to the g1006 mill. I got it many, many years ago when I wasn't aware of the disadvantages of a round column.

I'd been eyeing cnc mills for a while and after thinking long and hard at the pros/cons of a round column conversion vs buying a g0704 or rf45 as well as ballscrews I decided that for my needs right now I could get by on a very thrifty conversion. Just steppers where the handles used to be.

Total cost to convert both machines was around 1000$:
I'll get a better picture later, this was before I finished the lathe/mist coolant and got a proper table for the computer.

The backlash and quill are a pain but with some care I've been able to stay very accurate and I bought a 1/4" drill chuck to stay within the quill travel. I plan on doing a proper servo/ballscrew conversion on an IH001 or g0704 down the road.

I use visual mill for CAM.

I played with an HXT900 to get a feel for how strong they are and am quite confident they will work. I ended up stripping a gear by forcing it around by hand but they have quite a punch for something so small. They have replacement gear sets for 0.89$ so I'll grab a few with my next order.

The larger HK servos are really nice. The reviews mentioned how quiet they are but in person it is quite a difference. None of that annoying high pitched buzz. I'm not worried about their strength at all since they are like four times as large as the HXT900s.

10-24-2011, 11:45 PM
Time for some progress! I've finished up the first joint and figured out the hexapod is likely going to look.

The first set of pieces. The acrylic has been cut on a DIY laser cutter that I did myself based on plans from buildlog.net . It gives a great finish to the cut and works very fast. I'm also experimenting with engraving on the pieces.

The basic fit goes like this.

Mounting the servo into the leg is a tad more involved than one might expect since the wire at the bottom prevents it from being inserted. Its still easy enough to unscrew the top and push it in that way though.

And both servos mounted.

I played with a few ideas for mounting the servos. Initially I wanted to broach the spline straight into the aluminium but I realized how much work that would be and discovered it is very easy to press fit the servo horn into the aluminium. Works very well and looks identical from the outside.

Since the aluminium adds thickness to the horn I used longer screws to make sure the servo wont wiggle loose. The M2 hardware looks a lot nicer anyways.

The final (half) leg assembled. I tried to keep the wire straight by double sided taping it to the servo but not much seems to stick to the wire. I've found CA to work okay so once I pick up some thick CA ill do a proper job of it.

I kind of cheated here since I like avoiding anything that permanently joins two parts but I added some double sided foam tape to the back of the servo. It adds a lot of rigidity (2mm acrylic is like a wet noodle) and the leg feels (and looks IMO) really solid.

And the back. I hooked up the servo and things look pretty good. There is a small amount of play which is too be expected since the joint isn't boxed but the servo is surprisingly strong given how much leverage it has to fight and given that the weight will be spread out over at least three legs there should be almost zero flex.

The next three DOF will have bearings on the other side which will be much nicer. The bearings will all be press fits.

10-26-2011, 09:54 PM
Here are the parts for the femur, front and back:
Bearings press into one side wihle the servo horns fit into the other. Unfortunately the larger servos did not come with the one-sided horns so I had to cut them down.

The other joints of the leg also press onto the bearings. Gives a nice fit.



And with the screws all in. I used some CA to hold the wires in place. I'll finish up the acrylic tomorrow. Everything moves very smoothly for now. I sure hope I never have to replace any servos because those wires are going to be a huge PITA. The length isn't enough so I will have to extend a lot of wires to get everything to the board.

10-26-2011, 10:37 PM
Looking very cool :)

The aluminum seems like it would be a bit heavy, but will be interesting to see.

I think you electronics cost estimate of $125 is a bit low. $85 for the controller + servo controller... and then probably like $80 for two xbees and two xbee adapter boards.

10-27-2011, 02:49 PM
Weight might be an issue for the HXT900 servos but I'm hoping things will be okay. The construction looks pretty heavy but most of the aluminium has been milled down to 2mm or thinner.

Parallax has a good deal on XBees and a lot of money can be saved since the maple is 3.3v and no level shifter is required: http://www.parallax.com/tabid/768/txtSearch/xbee/List/0/SortField/4/Default.aspx
Its about 22$ for an xbee + breakout.

I'm not entirely sure how I am going to control the hexapod I will probably be using a laptop with an xbee. They sell adaptors that allow the use of an RC transmitter as a joystick which would be nice since that will provide 7 channels with my DX7. The alternative is skip the laptop and use an arduino or something to read the trainer port and send that through the xbee directly. I'll add those costs in once I figure out what is needed.

10-27-2011, 06:48 PM
And now to finish up the leg:
Laser cut acrylic pieces.

I decided to mount the coaxa servo facing outwards to shorten the useful span of the leg. The coaxa is already quite long since it consists of two servos (and the femur is comparatively short as it has no servos) so I want to hide some of that extra length inside the hexapod.

I was sure to leave a lot of wire around the joints to get full range of motion.


I've played around with it and everything looks good. The biggest question is whether or not the HXT900 at the end can support enough weight and it looks very good. It can lift the leg up from a horizontal position which means it is fighting around 6" of leverage. I'm planning on keeping it vertical for most of the gait so it should be under much less pressure most of the time.

Now I need to make six of these guys and design the body. While CNC does a fair bit of the heavy lifting I still need to be around for toolchanges and lining things up so its by no means a free ride. I will also need to tap sixty blind M2x0.4 holes by hand.

10-30-2011, 04:36 PM
I always get annoyed when I am reading a build log and need to wait months to see the finished product so I've been cheating a bit on my buildlog and started it a fair while after the project started.

Acrylic on the laser cutter.

All the parts ready for assembly. Probably upwards for 200 toolchanges and 100 hand tapped holes there.

You might notice there are 7 legs here. I made a full test leg before starting this buildlog and made some changes to the parts. The 6 new legs all have the same toolpath so the finish is perfectly consistent.
Legs are about 10" long extended.

All the wires glued.

Arranged approximately how they will be in the final hexapod.

It took about three weeks to make all 6 (+ 1 test) legs and I'd guess around 40 hours including all the design work.

There is a fair deal to think about here. I need to decide how to best do the wiring and what the body is going to look like. I know it will be circular but I'm trying to figure out how its going to look in terms of materials. I bought the sheet of carbon fiber but as there is none in the legs I'm not sure a carbon fiber body will look that good.

A lot to think about electronics wise as well. Making everything fit without binding will be a challenge.

10-31-2011, 11:35 PM
Now that the legs are done its time to work on the body. The idea behind it is essentially two rings, one to hold the bearings on the bottom of the legs and the other to hold the servo horns.





I will be making a ring for the top that is just for looks. All three disks will have a 2mm acrylic circle in the center to hold electronics and such.

I've been looking at alternatives but I've come to the conclusion that my only real option for keeping things looking neat is to individually resize each servo wire. Not looking forward to it.

11-01-2011, 10:19 PM
I finished off the topmost ring and cut a disk of acrylic for the top. To celebrate the robot at least looking finished for the first time I took it outside for some decent photos.




Don't be deceived, there is still a lot more to be done (I'd guess I'm halfway there). Fitting all the electronics into this guy in a visually appealing way is going to be extremely difficult especially since I will be making my own PCBs for the very first time.

Since there is so much acrylic going on I'm going to have to pay a lot of attention or it will look like a disaster.

One of my biggest obstacles will be mounting the micro-controller. I don't really want to solder to it since I will probably need several tries at the board but header pins add almost half an inch to the height which I really don't have available.

Anyone know any way to shorten these guys down?

I've also decided, for better or for worse, that my hexapod is severely lacking in coloured lights so I'm going to try adding LEDs in a classy way.

11-02-2011, 05:44 PM
Wow man. Pretty awesome how it went from talking about the idea to a assembled bot (sans electronics) in about a week. So far it looks really cool, looks like a ton of aluminum though, are you sure those servos will hold it up? How large is the bot overall, I dont know how big those servos are, so I don't know what to compare it to. Awesome looking so far man, lets see it moving in under a week too, now that would be something.

11-02-2011, 06:51 PM
I did cheat a bit, I had about two weeks of work completed when I started this build log so idea to assembly was about three weeks of work.

The servos are the real question here since my goal was keeping it cheap. If they don't have the power I will have spent a lot of hours on an expensive paperweight. That being said, I am very aware of that risk so I've been testing the legs and servos at every opportunity and they look good. It will definitely walk using a wave gait (keeping 4 or 5 feet on the ground) and should be okay with only 3 legs on the ground.

The blue servos are 9g (Similar to HS555) and the orange ones are 38g.

I'm hoping to walk within two weeks but electronics always seem to look great and then fry at the last second...

11-02-2011, 10:09 PM
Another little update while I push off the main electronics by finding other things to do.

I mentioned I wanted light but I'm not fond of just drilling holes for LEDs since it can look tacky. I decided on a colour changing acrylic ring around the body.

There isn't very much space if I want to keep the wires hidden so I used 36 gauge wire.

The servo colors: orange and blue, seem to have become the colour scheme of this guy so that's what I'm using for lights.

The plan is for both sets to be independently controlled via PWM.


On the hexapod.

I'm planning on having a current sensor in the finished robot which is basically a load sensor (as loaded motors will draw more current). I will probably make the lights change from blue to orange as more current gets drawn which I imagine as being really cool. Think of pushing on the hexapod and having it change colors or getting to watch the load change as it is walking/posing.

I imagine the colour changes will coordinate with the surface it is walking on as well as the movements very well which will just be awesome.

11-03-2011, 01:51 AM

I've been following your thread for a while now. And I must say I'm deeply impressed by your work. Excellent design and craftsmanship! I love all the details in your design.
The servos has been mentioned and they are also my biggest concern in this project. I'm quoting Matt Denton: "Buy cheap, buy twice".
The key to success using these servos is smooth and rather slow movement.

What's the overall weight of the bot?

Anyway, keep posting! Very fun to follow your work.

11-03-2011, 02:20 PM
Thanks, your video of morphex is what actually inspired this build.

My first drawings had the expanding body but mechanically it was a little ambitious so I settled for 4DOF instead. The round shape of my hexapod is entirely attributable to the morphex though.

I don't mind moving slowly as I am hoping for a "creepy" or at least living kind of feel to the movement. I want everything to be very smooth and deliberate. The biggest hurdle there will be the backlash and poor resolution of the servos but I'm hoping that with all 24 moving at the same time any inaccuracy will be hidden.

I'll weigh it later today. I've been meaning to for a while but it keeps slipping my mind.

I've been reading up on hexapod gaits (and how you handled the fourth degree of freedom on the t-hex) and I'm hoping to implement something that allows arbitrary motion combined with any stance.

I've got a few ideas in terms of how the IK will be handled. I want to implement a restriction based algorithm for robustness but I also need to the gait to look coordinated so I will be inputting "fake" restrictions based on a timer that will put out a normal gait.

The idea is that in regular motion the timed restrictions will be the only triggers for leg swings and the controller will emulate the preprogrammed gait. If a sudden direction change or strange stance causes the legs to go out of sync the original restriction algorithm will kick in and fix it.

The preprogrammed gait will probably be a wave gait as described here:
http://books.google.ca/books?id=EfX0miZllggC&lpg=PA14&ots=HAyOAyC2xw&dq=insect gaits&pg=PA13#v=onepage&q=insect gaits&f=false
I like it because it seamlessly transitions from only one leg off the ground at a time to the full alternating tripod gait so any speed changes will be continous and smooth.

11-03-2011, 10:05 PM
Weights are as follows:
Legs: 1332g
Body: 530g (Aprox)
Batteries: 121g

Total: 1983g
Where the servos weigh 564g of the total.

I ended up taking a few tries at the bottom plate for the hexapod:

This shows the basic idea for how the batteries will be held and how I plan on mounting an LED to the base. I didn't really like how the heatsink sticks out and the fit/finish was kinda weak so I tossed that one.

Second try:
By some marvellous stroke of luck, the acrylic is sized exactly perfectly for a press fit of the batteries. The batteries can be inserted and removed without force but will not fall out on their own. If you look very closely the bottom is actually made of two acrylic disks, this should make it easier to mount the electronics.

And here is the LED running at around half power.

With the top on. Looks kinda like a fancy lamp I guess. The bottom LED will be saved for special occasions since leaving it on constantly would be pretty obnoxious.

After going through the entire process a second time I realized I'll need to redo the base yet again since I have no mounting holes for electronics.
I was going to just tape them in but I've put so much time in already I should probably do this right.

11-04-2011, 05:57 PM
I finally redid the upper acrylic plate and mounted all the electronics. The boards without holes were attached to a styrene sheet. In general, when building I try to make it at least possible to disassemble and reassemble so all the wires are attached to screw terminals on at least one end.

Right now the regulator, servo control board and current sensor are mounted.




I used a few drops of glue to keep the wires where I want them.

You might notice that the USB port on the servo controller is in an inconvenient spot because a leg will block it. I want to control access to the pololu servo control because it actually has some pretty nifty features such as a GUI and posing/scripting system that will save me some development time. I considered moving the usb port but I didn't want to make any permanent changes to the board so instead I cut up a mini usb cable:


Next up is mounting the legs. I will be individually resizing each servo lead... Let the splicing begin!

11-05-2011, 11:10 PM
And here we are. Individually spliced each wire and heatshrinked the join.





I got three done today and the legs work okay. The servos have pretty shitty tracking and some of them have some trouble centering. It stands fine though so I think if I'm careful with how they move it will be fine.


I considered sleeving the wires but the distance is really short and the connectors are large so IMO it doesn't look that great. I'll do the remaining legs tomorrow.

11-05-2011, 11:29 PM
Very professional construction, looks quite nice. Can't wait to see it all together!

11-06-2011, 03:55 PM
Got in all in. Was pretty gruelling but worth it in the end.

I ended up just using zip ties on the wires.

Since the leg interferes with the USB port I used my cut up cable.

Standing in a neural position.

Standing high.


Performance wise it is about what one might expect. Standing the entire hexapod consumes less than 3 amps which for reference is less than the peak consumption of a single AX-18 servo. It has enough power to stand on four legs but 3 legs don't have the strength.

11-07-2011, 03:19 AM
Looking good so far!
Did you measure the total weight at this point? I would say that a current consumption of 2,6 amps when standing is rather high for a hex of that size. That might have to do with the servos working hard. My T-Hex draw 1,59 amps when standing, Phoenix draw less than 1 amps.

Btw, did you connect the two LiPo's in parallel?

11-07-2011, 10:07 AM
The total weight including the batteries and electronics is almost exactly 2Kg. I suspect the high current consumption is due to a combination of weight and cheap servos.

The two batteries are in parallel. I was running the servos at 5V in this photo but decided to increase that to 6V for the extra power.

For these photos I arranged the legs by eye which means some are supporting much more weight than the others. When I individually measured the servo temperatures some were at ~20c while others were at 40c. Hopefully when I check all the angles with a protractor things will be more smooth.
At this point the weight is clearly an issue but I don't think the problem will be bad enough to warrent redesigning the hexapod. It should be able to walk and I like how it looks.

At least this gives me an excuse to make a second hexapod with everything I've learned.

11-07-2011, 10:21 AM
It'll probably also draw significantly less power if you make the body higher off the ground by changing the leg positions. Right now you've got a fairly wide stance there. Still, very cool :)

(this inspires me to get around to measuring how much current my Quad takes to stand in place.)

11-07-2011, 06:57 PM
I've been doing some electronics work and since its not my strength its been slow. I've been carefully testing everything on the breadboard to make sure there are no surprises.


I'm happy to report that everything seems to work as advertised and I can even monitor battery voltage and current draw.

Anyways since there isn't really much going on I decided it was about time I took some decent photos with the lights and electronics in.


Next I'll be etching PCBs for the first time. Should be fun.

All those wires and a fair deal of stuff still need to fit inside. It's looking like there will be less than a millimeter of clearance.

11-08-2011, 04:52 AM
It'll probably also draw significantly less power if you make the body higher off the ground by changing the leg positions. Right now you've got a fairly wide stance there.

Very good point, It looks like the horizontal distance between the femur joint and the tars (foot) is relative large. Also, like pointed out, correct calibration is important too.

I like your latest pictures, very cool lighting effects.

Good luck with the electronic board.

11-08-2011, 10:36 PM
I'm doing toner transfer with a HCl+H2O2 etch for the boards. It was fairly easy although the quality of the boards is only okay.


My xbee breakout boards have not arrived yet so I needed to make something to work with the 2mm pin spacing.
I cut up a 0.1" header and used the pins from it. They are very close but do not touch.

I was able to test and set up the xbees and they work pretty much as advertised. I was very surprised at the latency they introduce though. I have pro modules from a few years ago so I feel okay about putting one inside the hexapod.

Anyways, this is the intermediate board I came up with to help me fit everything in the hexapod. I really wanted to avoid permanently soldering the expensive chips for when I fry something. Headers take up a lot of room...

I'm using some blue fiberglass which is pretty snazzy. That is the reason why I am not doing isolation routing on my mill.


I'm powering the xbee off the maple minis internal 3.3v regulator since it was easiest and had the capacity.

Drops right in.

The idea is to slide my "stack" of boards into a main PCB that does the heavy lifting:

I'm hoping the exposed electronics will look cool.

11-10-2011, 06:13 PM



I made the board but it doesn't look quite how I hoped. I'm sure with a proper silkscreeened double sided board and surface mount components it would look pretty cool but I can't shake the feeling I'm just showing off the part of my hexapod I am the least proud of...

Anyways, for better or for worse the board has a few mistakes so I will have to redo it. I'll try to make it smaller and look nicer.

11-11-2011, 12:20 AM
See if the board you made will do what you want it to. Fix the issues with jumpers.
Your going say to yourself man I wish I added this will I was making the board. So your probably going to make it over more than once

11-11-2011, 07:09 PM
I messed up and used npn instead of pnp for some of the transistors in the original board. Luckily since I am etching the boards myself and it connects via header pins it is very easy to swap them out. I've done this on the breadboard quite a few times so hopefully it has everything I need.

I was originally doing toner transfer from my laser printer and had been having a little trouble. On a whim I tried photocopying the design on a very old photocopier and the toner worked way, way better. I was able to halve trace width/spacing and got a much cleaner board:
I like it way better than the old board.



Now I just need to shorten some wires and cut the plate it will be mounted on. Everything looks good on the bench.

11-12-2011, 08:16 PM
Finally all that remains is to mount the board nicely:
I laser cut a disk and the board was attached with double sided tape. Then I just had to shorten some wires and put in the connectors.





A bit messy but overall I'm pleased with how things turned out.


I'll throw on 175$ for all the random stuff I ended up using (Wires, PCB, Connectors, hardware... that stuff adds up fast...) taking the total build to aproximately 500$.

If/When I do this again I will focus a lot more on weight and use stronger servos. HK has a line of high torque servos with metal gearing that has the same form factor but twice the power of the servos I am currently using. They cost about 10$ a piece.

This essentially marks the end of my build. Everything has been tested and all the features work - all that remains is code. I'm burning out a bit and I will be returning to school in January which means I have only a month and a half to finish the mechanical portion of my projects (otherwise I'll be stuck with a file and hand drill).

If I get some spare time I'll try to get it walking and do the rest of the code after christmas. I'll try to grab some better photos as well.

It might have looked like I knew what I was doing in the build log but I assure you that it was not the case. I have a fair pile of scrapped pieces and even more that were thrown into the trash or destroyed in disgust.
In the end I learned a lot and it was all worthwhile though.

My next project will involve four of these and a nice stash of carbon fiber:

11-12-2011, 08:55 PM
You have entirely too many fun tools and toys.. Good work.


11-13-2011, 12:01 PM
That looks like the rear blades off my Xcopter 450 :p

11-17-2011, 03:16 PM
Hexapod Pictures: Winter Edition!