Parallel Leg Mechanism Design Discussion

[Tyberius offline]

Alright, so after finally seeing this type of design in action at the Korean Robot Game Festival, I've started giving it some serious thought. I have an interesting email discussion forming with some colleagues of mine and I'd like to share that as well as open it up here for some Q&A. Consider this an open discussion on furthering humanoid leg designs, as I feel this will be beneficial to all humanoid builders and especially those of us building bipeds for Mech Warfare.

What is a Parallel Leg Mechanism (PLM)? I'll get to this shortly, first let's lay down some ground for comparison.

In a standard humanoid leg design, you generally have 3 pitch servos: ankle-pitch, knee pitch, and hip pitch. For the sake of clarity in this discussion, we're going to assume that hip-roll and ankle-roll also exist, giving us a minimum of 5dof (or 6dof if you have hip-yaw, but that's not important for this discussion) for a starting point as a 'standard leg' design.

Here's an example of a standard (non-PLM) 6dof leg design:



So with that said, let's ignore roll & yaw servos and focus on the pitch servos, as that is what is important here. Each one of these pitch servos gives flexibility to the leg, but the leg is only as strong in terms of lift/push/pull as a single servo joint. Each pitch axis also suffers from gear backlash, also known as 'gear slop'. These two problems end up hurting biped leg design a pretty decent amount, as your payload is always pretty limited, and the more weight you put on a robot, the more backlash hurts you.

How about ways to prevent this and/or increase torque? You could very well double up each pitch axis by placing the servos back to back. This would effectively double your torque and could be used to eliminate backlash. Problem here is you now have 6 pitch servos and a very wide leg, not to mention the increased weight. The weight then starts to cause a problem for your ankle and hip roll servos. So while that's one option to increase your torque and help eliminate backlash, it's probably not ideal for more compact robots.

This brings us to Parallel Leg Mechanism (PLM) Design. What is it? In short, it uses mechanically paired linkages to bind your rotation points together. This effectively creates a parallelogram mechanism for each the top and bottom leg halves. There are a couple different variations of it that use a different number of pitch servos. Here is a pretty standard example of it using 2 pitch servos, one in the ankle and one in the hip:



Here's another variation that uses 2 in the knees for each top/bottom part of the PLM and then normal pitch servos in the ankle/hip:



And some video of the bottom half of the leg:

phpBB [media]

Another variation uses a total of 4 pitch servos, 2 paired per PLM in the upper and lower leg. Here is a very rough draft of a model I was playing with in Inventor:



Another example of the 4 pitch servo PLM can be found at these links:

http://bipedrobotnewsjapan.blogspot.com/2009/01/mercury-new-type-robot-leg-parts-for.html

http://www.rakuten.ne.jp/gold/grass-road/melissa/melissa.htm (scroll down a bit, you'll see it)

And... http://www.rakuten.co.jp/grass-road/507141/852661/926782/

CHARLI from RoMeLa: http://www.vt.edu/spotlight/innovation/ ... harli.html

And lastly, here's a great japanese blog of 'Ignus' - the red robot shown in the videos: http://snjrobots.dtiblog.com/

So as you can see, a PLM design leg uses additional linkages which bind the rotation of the axis points together mechanically. The ankle, knee, and hip all remain parallel to the ground.

Pros:

• Ability to utilize 2, 4, 8, 16 pitch servos per leg.
• Increased stability/reduced backlash.
• Increased torque when servos are added to PLM, without widening leg.
• Can be very compact.
• Ability to only use 2 pitch servos instead of 3 while maintaining a functional leg, reduced weight.

Cons:

• Decreased range of motion. (can't kick, etc)
• Standard IK models will not work without additional constraints.
• More mechanically complex.
• Potential to add stress to servos if not properly paired.

So what does this mean for you (I'm looking at you Mech Warfare biped pilots)?

If you utilize a 2 pitch servo PLM leg design, you've eliminated a 3rd pitch servo and thus lowered your weight while severely reducing your backlash. This means more payload and stability.

If you utilize a 4 pitch servo PLM leg design, you've greatly increased the torque on each PLM and have the ability to virtually eliminate all backlash. This means more even more payload and stability.

I'll dive into the design implementation a bit later, there are definitely some pitfalls and tricks you need to know to successfully pull it off. I'm working on upgrading Giger to a 4 servo-pitch PLM design as well as tinkering with a 2 servo-pitch and 4 servo-pitch AX-12 biped for mech warfare, so I'll have plenty to share in the near future.

Here's a video that demonstrates the difference between a standard and PLM leg design. The first 20 seconds are using 4 pitch servos in a standard, non PLM, configuration (so double knee servos). After that, you see the same robot with a redesigned 2 pitch servo PLM configuration. Look at how much smoother and stable it is. This doesn't even demonstrate the increased torque capabilities of 4+ pitch servos on a PLM design.

phpBB [media]

So that's about it for now. Thoughts, questions, comments?

I'm going to include the emails of us kind of thinking through all of this (we're all still learning) thus far for those of you interested in reading more of the discussion, but for those of you who've had enough of my yapping, feel free to stop talking here!

I have two designs in my head right now.

First Design:

Uses 2 pitch servos per PLM, for 4 total in the leg dedicated towards pitch.
1 hip-pitch servo which is mechanically tied to a 1 upper-knee-pitch servo
via the PLM, and then 1 ankle-pitch servo which is mechanically tied to 1
lower-knee-pitch servo via the PLM. The drawback to this is that you limit
your range of motion significantly as you can't really 'kick'. The hip,
knee, and ankle more or less stay parallel to the ground. The bonus to it is
that it could theoretically double your pitch torque while nearly
eliminating backlash in the pitch axis's of the leg. I think without the PLM
tying the servos together (such as a normal leg design but with double knee
servos), you'd just get double speed. Am I correct here?



The first picture demonstrates this using only 2 servos, my proposed design
would replace the idle hubs in the knee with servos on the previously
un-powered linkage from each PLM, so that each linkage of the PLM has a
powered servo on one side of it (ankle-knee, knee-hip). The 2nd pic is my 4
pitch servo version.

http://img214.imageshack.us/img214/8765/dscn0049z.jpg (2 pitch servo
version)

http://forums.trossenrobotics.com/gallery/files/1/4/9/2/legprototype_original.jpg
(4 pitch servo version, poorly designed rough draft)



Second Design:

Now the second design idea utilizes a normal pitch servo in the ankle and
hip, and then two servos in the knees- one for each PLM (upper and lower)..
So instead of having two servos per PLM, you only have one (each located in
the knee). This doesn't sacrifice your range of motion, but you lose your
backlash reduction in the ankle/hip. I think you end up with less backlash
overall though, because the knees have the PLM supporting them to the
ankle/hip. You don't get the (questionable?) added torque benefit along all
pitch axis though.

This design is demonstrated well here:

http://24.dtiblog.com/s/snjrobots/file/CIMG3338.jpg

The four advantages I see for a PLM design are..

It's just easier to make a stiff frame.
Lets you move an ankle servo to the knee.
Lets you double up your servos in a convenient and compact way.
You can get away with 4 servo legs if you're fine with your foot
always parallel.

The first design with the 2 servos only benefits from the stiffer PLM
frame and less required servos.

I think the second design with 4 servos only gives you a stiffer PLM
frame. The geometry you have setup does not allow for doubled torque
since paired servos are not in parallel and it also does not allow for
doubled speed since the paired servos are not in series either. It
would allow for backlash tweaking though.

The Ignis design puts the PLM servos in the knee so there are more
options for the ankles. I guess he does not like the parallel foot
issue and that's the point of the extra ankle servo. I have no clue of
the point of the extra hip servo.

I think the biggest bang for the buck comes from putting all the PLM
servos in the knee and doubling them up there too. This gives you
doubled up torque and lots of room in the ankle and hip. My current
MCAD stuff is experimenting with 4 servos in the knee ( two for each
half ) and pushrod ankles and hips.

A 4 pitch-servo PLM Leg, if done correctly, should increase torque. It's never going to exactly double it, due to friction and magical fairies, but it's certainly a load distribution method.

Comparison between the two types of leg designs, PLM and standard:

Lets say we have a standard, non parallel leg design with double knee servos, so 4 pitch-servos total but without any mechanical link between them. If two of the servos move in a counter rotation of 30 degrees for X seconds, they'll have covered 60 degrees in X seconds. This can create double speed, but not any additional torque as each pitch axis is only as strong as a single servo. You also still have backlash to contend with.

A PLM leg design with 4 servos ties each top and bottom pair of servos together mechanically, so that the servos cannot move in a counter rotation and must move at the same speed. If say, the two top PLM servos are paired via software or the EX-106+ link cable, they effectively move as one servo and the mechanism helps distribute the load between the two. If each servo moves in a synchronous rotation of 30 degrees in X seconds, they'll have covered 30 degrees in X seconds.

A single PLM assembly (top or bottom half of a leg) would have 8 rotational pivots, all mechanically tied together. At this point, it doesn't really matter which of the pivots are powered, does it?

[mog123 offline]

I really like the PLM design, I first saw it in OmniZero.4 and then in IGNIS. Can you explain how the servos in OmniZero.4 are aligned? I can;t actually see anything from the pictures.

http://www.1mm.jp/m/robo2007.html

[UncleBob offline]

My second fight opponent was a PLM design. It is huge but still under 2KG due to PML design.

People say it can't Kick but the Japanese robot played football, no ?

Main advantage of PLM is that it can increase stabiity and reduce cost of 2 servos and reduce weight.

Right now I am playing around with other dirty method to stiffen up the angle servos with rubber band and wrist band.

[27Loco offline]

Yes, looks like it kicked quite good in the video. This robot is using this design too.

phpBB [media]

I wonder if you can do this with a Bioloid Premium too.

[UncleBob offline]

actually i thought about cutting some plastic and do the leg that way and remove the servos. this takes 120grams away. Then you can put in an RX64

[UncleBob offline]

what cad app can i use to do this ?

phpBB [media]

[mog123 offline]

Robita-kun used Alibre design for designing ignis. I'd recommend using Autodesk Inventor. It's easier.

[FabinovX offline]

you can easily do such a thing using Dassault CATIA v5.

[Tyberius offline]

Some of my initial cad work towards a PLM leg design:



Very "Melissa" inspired.

[Tyberius offline]

Some initial progress on the next revision of my robot Giger using PLM:

[27Loco offline]

Unbelievable how good this thing looks.

Looks like nights of programming

[veltrop offline]

Tyberius I love your Giger design. About how tall is it?

[Tyberius offline]

Thanks!

In it's current proposed form it would be about 80cm when standing fully upright. This is including the torso redesign that I have in the works.

[xanhast offline]

I like the design

That running bot and IGNIS are both really impressive.

Not sure if this is a given or if someones suggested it already but couldn't you configure the parallel servos to use the same signal wire?

For bus servos, use the same servo ID (don't know if that's possible)

[Tyberius offline]

Nice thing about the EX-106+ is that they actually have a separate link cable feature that does exactly that. You can pair the servos together using it and then just send commands to the 'master' servo.