A Multi-Purpose Hexapod

[Th232]

I figured I might as well start a project thread on the monstrosity that I'm planning to build. First, here's a screenshot of the current design in Inventor:



In that pose it's 550 mm long, 718 mm wide and 337 mm high. Weight is just short of 10 kg, and will exceed that since I may add more batteries and sensors.

Overall goals:
* Make a good sized robot. 'Nuf said.
* Put my mechatronics engineering degree into practice. We also didn't learn as much in some areas as I hoped we would.
* Modularity for added functionality and ease of repairs/replacements. I have thoughts on a 6 DOF arm, a filament extruder and more. Did think about a laser cutter, but that would need a lot of planning and safety considerations. Probably won't do it.
* Allow for autonomous activity or manual control.

Breaking it down a bit further:

Mechanical goals:
* Stable movement over a wide variety of terrain, including stairs. Using this outdoors would be nice, but could get very interesting...
* Enough grunt to carry a multiple kg payload (see modularity goal) at a reasonable pace
* Easy to replace legs if so required.
* Reduce strain on hip servos, especially Hip-Pitch and Hip-Yaw
* Mounting points for additional modules. You can see the front one in the above image, 4 holes on the L-brackets at the front. Subject to change, but probably won't differ by much.

Electrical/Electronic goals:
* Battery power to last at least 3 hours. This may be a pipe dream.
* Power management board. Seems like all protection circuitry is made for relatively low power applications (e.g. current draw limited to <20 A). Design for a custom one is going well, a problem will probably crop up though.
* Recharging station. So far, this is much more annoying than it sounds.
* Onboard brains. Eyeballing a FriendlyARM, Beagleboard or similar single-board computer. Will most likely need secondary boards to interface with sensors. Those will be AVR based.
* Sensor suite mostly undefined yet. Only thing I've designed into it are pressure sensors on the feet, and an IMU is practically guaranteed. Will most likely include a vision system, rangefinders and so on.
* Wireless communication. Most likely going to go for Bluetooth or WiFi. From memory the data transmission rates from XBees will be too low to stream video (may be wrong on this though). More and more single-board computers have WiFi though.

Software goals:
* Probably going to go for 3 DOF IK straight off the bat. Constrain Hip-Pitch servo so that the plane generated by the femur and tibia is normal to the ground, then figure out algorithms to use rotation in that plane later. Hello Denavit-Hartenberg convention...
* Would be neat to have some code to read G-code and transform it into robot movements. That'll happen after I make the filament extruder though.

Process plan is:
* Get all servos and mechanical components.
* Build chassis, leg joints &c.
* Build power management board, signal isolaters and whatever else I need.
* Buy single-board computer.
* Get computer to control servos
* Implement IK and gaits
* Make final decision on sensor suite
???

This will be very long, and probably pricy in terms of my wallet and sanity. At least I have a mill & lathe to work out the mechanical side of things.

[Th232]

So a British fighter pilot in WWII was shot down over Germany and was captured by the Nazis on the ground. He was pretty injured, and they had to amputate his leg, so he begged them "Please, if you have to cut my leg off, can you drop it over my country next time you send a bombing mission so it can be buried there?"

They decided it was a fair request, so they dropped it on their next raid.

A week later his other leg had to be cut off due to his injuries, and again, he asked them to drop it over England on the next raid, and again they did so.

The next week they had to cut off his arm for the same reason. Again, he asked the Germans to have it dropped over England on the next raid. The German barked at him "Nein!"

The pilot was confused, and asked "Why not? You've done it before!"

The guard sharply replied: "Because we think you're trying to escape!"

Why do I say this? I think one of Giger's legs is in Australia...

[Suicidal.Banana]

Haha nice that cant be right

[Tyberius]

So a British fighter pilot in WWII was shot down over Germany and was captured by the Nazis on the ground. He was pretty injured, and they had to amputate his leg, so he begged them "Please, if you have to cut my leg off, can you drop it over my country next time you send a bombing mission so it can be buried there?"

They decided it was a fair request, so they dropped it on their next raid.

A week later his other leg had to be cut off due to his injuries, and again, he asked them to drop it over England on the next raid, and again they did so.

The next week they had to cut off his arm for the same reason. Again, he asked the Germans to have it dropped over England on the next raid. The German barked at him "Nein!"

The pilot was confused, and asked "Why not? You've done it before!"

The guard sharply replied: "Because we think you're trying to escape!"

Why do I say this? I think one of Giger's legs is in Australia...

Attachment 3959

Haha, easier to ship when the ankles are intact. Figured you might want to experiment with some other sized brackets.

[Th232]

So... uhh... updating the firmware on two of them (daisychained together). How long is it meant to take? It's been at 0% for 5 minutes so far. There's a small alarm going off in the back of my head and I'm starting to listen to it...

Edit: Well, if anything they're fault tolerant. Updating was still at 0% after half an hour, I pulled the plug and closed the program. No ill effects, servos still working. I'll try to avoid updating the firmware unless I absolutely have to. Wonder why it didn't work though. Might be firewall issues.

[cire]

So... uhh... updating the firmware on two of them (daisychained together). How long is it meant to take? It's been at 0% for 5 minutes so far. There's a small alarm going off in the back of my head and I'm starting to listen to it...

Edit: Well, if anything they're fault tolerant. Updating was still at 0% after half an hour, I pulled the plug and closed the program. No ill effects, servos still working. I'll try to avoid updating the firmware unless I absolutely have to. Wonder why it didn't work though. Might be firewall issues.

Did you try it with only one servo connected? I am pretty sure you are not supposed to update two at once, although I have never used the EX series.

[Th232]

Just tested it, no joy with just one connected either. Same thing, just says it's at 0% the whole time.

[Th232]

Design is slow when you've got a PhD to do. Playing with Giger's leg, it was tempting to rework this into a 5 DOF quad (3 DOF leg + 2 DOF ankle). Very tempting. Given the modularity of my frame, I still might. But that'd be later on.



Major differences so far:

* Simplified and lightened hip joints.
* Relocated batteries to underside of robot. Fire resistance reduced, but hopefully a good combination of thermal sensors and power management will help, along with a quick release system.
* 5 DOF robotic arm. I wouldn't mind trying the MX-106s for the arm if they're out by the time I make this. Power may be an issue though, given the difference between the EX and MX series.

The reason for the gaps in the legs is that there aren't models for the OF-106HXL brackets, so I'm just using the regular length ones and adding an offset.

To add:
* Lock mechanisms for legs and arm. It's 11.5 kg, 60 cm long and a bit over 75 cm wide in that pose. Carrying it with the legs dangling everywhere and the power off sounds like a recipe for trouble.
* Sensors...
* Cargo pallet on the back or similar. A quick release system would be nice, but not essential.
* Quick-release tray for Lipos, in the event something does go horribly wrong.

[Th232]

Big changes, both in the robot and in the world. Markets, global economy and so on have gone downhill. In particular the AUD:USD exchange rate has dropped by 8 cents in the last month. This project is continuing, just that purchasing might happen at a slower rate (Tyberius, I'm still good to send payment this coming week).



Changes in the robot:
* Changed from CF rods for the legs to sheet metal. This'll be more expensive, but I think it'll be worth it. Weight penalty is another issue. For those who are curious, the legs are about 50 mm wide and made from 2 mm thick Al.
* I'm leaving room to expand the legs to 5 DoF, although that'll probably only happen in the very far future or if I win the lottery (and I don't gamble ).
* I made a minor tweak in the skeleton width so that I can have the legs fold in like those on the right (far) side. I'm tempted to have the locking system as just a bar or latch that I move manually.
* Quick release tray for batteries. With good enough torsion springs it should survive bouncing about, but still release the tray when I pull on it.
* Added a yaw+pitch mounting for sensors on the head.

On that note, I did some rough calculations. With all 6 legs on the ground (i.e. standing still) and assuming the servos are powered at 14.8 V and that torque and current are linearly related, it should it should draw ~27A when it's just standing still. Motors are well within the operating range, but the implications for battery life are obvious. Any thoughts or comments? I've got some thoughts on the software side of things (have it sit down if it's not doing anything), but I can't wait until I build a leg and test it under load to see what the actual current draw is...

I dumped the arm, going to keep the robot simple for now. So with that I just need to finish off the locking system for the legs and work out the feet. Then time to hammer out the electronics.

* Sensors: At the very least it'll have pressure sensor on the feet, and a video camera + lighting and some kind of rangefinder on the head. Home-made LIDAR would be nice. I've read through the LIDAR discussion on the forums, so this'll be one of the major challenges after I've gotten the robot moving. On that note...
* Plan is to use a fitPC, although I'll offload some of the more basic sensor and motor stuff to an AVR or two.
* Power electronics. I'm planning to have a charging dock, which means on-board charging and balancing of the LiPo cells. Found a board (dunno if it's acceptable to link to other sites) that looks like it does everything except allow for the high current draw of this robot. Still need to buy it, but it doesn't look bad. If I do, hacks to be made:
** Motors and so on will be directly connected to the battery, otherwise I'll trip the current protection as soon as the robot tries to stand up. The LiPos should be able to handle the current draw but I might put some kind of current sensing device in, just in case. I could just change the sense resistor on the board but I'm not sure about what the FETs and are rated to.
** Find the signal that turns the discharge FET on/off. This will ideally be fed into an AVR then into the fitPC. If my robot is moving and the battery drops below the cutoff voltage I still want it to shut down quickly, but in a graceful manner rather than having all power suddenly cut (cue *thud* sound as the robot hits the ground). I'd also be very unimpressed if it shuts down half a metre away from the charging dock.

Any comments on how I'm going about all of this would be appreciated as well. Having a mechatronics degree only means I know enough to be dangerous, rather than do anything properly...

[jwatte]

3 hours battery life, and more than 20 A power draw. That's 600 Ah of batteries. I have a 10 Ah 12.8 V 4-cell LiFePO4 battery pack for driving my on-board computer (quad-core Intel -- need the CPU oomph :-) and it weighs at least a kilo. Multiply by sixty...
These are the cells: http://www.all-battery.com/32v10ahlifepo4lithiumironphosphaterechargeablebatt ery.aspx
This is the protection board (cuts out at 30A, which is the burstable limit for the cells): http://www.all-battery.com/protectioncircuitmodulefor3cellslifepo4batterypack-2.aspx
So far, that's working out swimmingly for me.
They also have "high-C LiPo cells" if you want to draw 100A: http://www.all-battery.com/37volt-10000mah10cli-polycell.aspx
(At $20, that's a 10 Ah cell at $2/Ah, which seems pretty sweet! I stay with the safety of LiFePO4 though)

Also: lucky you! They didn't have mechatronics degree programs when I was in school :-)