Bipeds

[doc222]

X-2.5 specs(For mechanical testing, no senors yet)

I need to be sure this think can do what I have planed, no need to have sensors if it cant walk or climb stairs. So first programs are are just to to prove mechanical ability. 5 dof legs are all most bipeds need. Asimo and about all other bipeds use 5 dof per leg. With Lego the issue is weight(power to weight ratio), compactness and scale.

Specs right now:
1 nxt
10 NXT servos
2 home made Lego La's( these are very powefull and seem to work well)
4 Mindsensors NXTmmx( these tings are the coolest thing since sliced bread)
4 Lego linear rams(2 at each foot for leaning) special note here Lego linear rams have a flaw that has been blogged at TechnicBricks(TB) and as we all know the last big kit was recalled. The issue is the metal screw that turns within the threaded sleeve. It wears the plastic sleeve very fast. Under loads less than when the internal clutch slips it has been seen that chips of plastic are been made, like saw dust. since its in the inside you wont see it, but after time the friction creats a rough movement and in time failure. They took a few apart to check this out at TB( note duing this ruins them as they are glued in assembly). I asked them if they noticed any lube in the Linear rams? They said no it was a dry assembly in the old ones. The new ones the screw part has been gapped to wider threads and alot of grease is used this is the fix. But it causes the new La's to be loose with more free play. My fix is just to lube the old one for now and X2 will prove if Lube is all they needed to start. This is my thoughts since anything mechanical with out oil/grease is not going to last in my lifes experiance( I have worked on hot rods all my life). We all know what happens if you do not have oil in your motor).

Doc

[mattallen37]

...We all know what happens if you do not have oil in your motor).

Doc

Yup, people assume it's a 2 stroke . I agree, it was very dumb of Lego to not grease them. I don't like the protection mechanism either. If it clicks, stopping linear motion, then the encoder on the driving motor will continue, so the problem is that you will loose your position count accuracy in a matter of a single click.

[doc222]

First baby steps of X-2.5, testing new Leaning functions.

http://www.youtube.com/watch?v=cxRAF2W3 ... r_embedded

[doc222]

After seeing how other more advanced Biped Robots can walk side ways. I programed X to try its legs at this. But during the first test runs i found a few tweeks were needed. Like the Knee ankles inteference, I found that If the legs are extended out side ways there is an issue with Hip motors will bind with the gantry of the hip lean assembly. Also that when the hip lean is at full tilt it will hit the systems contorls(mmx boxes). So I hade to move the system controls up one studd and relocats the hip motor. Like any work in progress things need to change to adapt. Each little step to improve the whole.

X taking its first side stepping moves. To step up on the leg stepping out it takes 7 motors at one time all must do what they need perfect or it will fall. I founf that this is the first dynamic move for X as well it catches its sefl and then moves on to this foot. Not completely dynmaic but not totally statics as well.

[doc222]

I happy with the results so far with X-2's leaning function's. Though it has been alot harder to plot compared to the older way X-2 did this with just shifting a mass side to side. Leaning at the ankles and hips brings in what I can best describe as the 3rd dimensional action. This also seems to change as angles change as well. So many little correction are going on at the same time. At times up to 8 motors are on call to stay in that "window of balance". This is a very basic program right now, once again just to prove concept and mechaical ability to do a task. Building this biped has been a work of years now(hard to believe), many failurs and designs that did not work or just not well. I have thoughts for improvments of this that may or may not improve, such is the way things go. Every change means checking again and reprogramming to see what changed.

So today X-2.5 has reached the point it has now climbed up steps. I have not even started the "can it go down steps" yet.

X-2.5 on a two step stair.

[aswin0]

Wow,

This is impressive. I really like the way the robot leans forward to take a step. Sorry I rang you up when you were shooting this movie.

How did you program the coordination, by trial and error? Or do you have some sort of model of the robot in your program that allows you to calculate the position of the center of gravity so that you can keep balance in some smarter way?

[doc222]

Right now its a dumb as I can make it even at that it seems to take up alot of memory to do even simple things like stairs. Not sure how smart it will ever be, some simple AI i hope but Bipeds that can climb steps and do other hard task are not to easy to build, in Lego even harder so much play in the gears and connections flex. Its going to be even harder to make it smart since these varibles can add up and well, gravity is always looking to bring it down. I have geared this almost as fast as i can make it and still be able to lift its own weight with one leg, from the servo first reduction is 2:1 at the planitary is 40:1. The faster you make a walker the more Inertia to over come. It gets so very complicated REAL fast. But that is the reason it keeps me interested over the years. If it was easy we would see many of these about, to date i think this is still the only Lego biped to climb anything at all ever. I think I will have have to gear down to 3:1 at the first reduction for the AI to keep up with gravity. Its work in progress that may take a life time to finish

I have a feeling to have it know where it is in space will take perhaps alot more programming then I know. It is the kind of thing that most projects have teams of guys working on. I am more of a mechanical designer and a padawan programmer. The only other Lego biped that is trying to use PID is taking along time and I wonder how he is doing with Legasimo? His has the NXT and batteries off the bot so weight is not an issue, mine is self contained and carries 12 aa's, it rather heavy. Just under 5 kilo's thats alot of mass to stop and start. So many varibles. If I can get to say: walk along and know when the stairs are in range line up to the steps, Know how high to step and far, climb the steps and know when the top of the landing is, then know when they landing ends, then decend the stairs and walk again would my dream Lego Biped. But thats a lofty dream since even the most advanced Bipeds are far and few(not remote controlled) I can only think of 1 Hondas asimo?

Doc "dreaming of Bipeds"

[doc222]

After testing X-2.5’s for its basic functions of walking, climbing small steps of 1.7 inch. I wanted to test it at the last versions 5” steps. X-2.5 is shorter than x-2.1-4. The femur and tibia are each 2 bricks shorter for a total of 4 bricks. This was a real issue when attempting this height. Since this new design has so little “perfecting” of its design I knew that it needs to gone over and compartmentalized. With the mainy issues such as motor in the way when in a high angle move, legs that needed to be a bit stiffer and the worst of all the areas was the hips. It took along time to come up with a design for this area. Getting legs to move side to side and forward to back with little play and to be able to hold the load. But I took some short cuts to get into testing. When you look at how wide the hips are with those “flying wings” that support the linear rams from the outside to push or pull the femur. It was a simple way to do this and my home made linear technic rams worked well, they are very powerful. But, they are huge, heavy and rather ugly as well. I wanted to have them inside the pelvic area but then this area would have to HUGE as well. The socket for the legs to the pelvic was 8 wide on the inside, this was one brick wider even though the linear rams were mounted to the Flying wing.

Changes to X-2.6
Made the femur and tibia each taller by 2 bricks total 4 for each leg. This makes it possible to have a H beam construction in both legs areas. since the legs are longer the motors now can be parallel to legs. this keeps them out of the way ever at the legs maximum angle.

Compartmentalized the foot, using far less parts and made much more compact. before the motors hung out past the foot print area. This again was done just to get it done. The feet have went through at least 4 rebuilds, each better, but I feel now its is at its best possible state.

The hips. This is the area needing the most compartmentalization of the bot. The old pelvic area was about 30 studs long, and 8 inside wide. This is built from old technic brick, stud-less and is a H beam. This area needs to very stiff, to the point of needing to be almost like a solid. The new pelvic is only 15 studs long and now 7 on the inside, H beamed and holds the legs and now uses Lego LA's inside and in-closed. Lubing these Lego LA's seems to have made them durable(time will tell and we will see if they last) if they don’t I will go to Firnelle LA’s in the hips. The design is cross ram, left ram moves right leg and left moves right. This has made the robot a lot lighter as well.
X-2.6
Side view


Front View


Foot compartmentalization


Foot OLD


Legs bottom,tibia H beam


Legs top Femur H beam


Hips/ Pelvic area
Cross ram( short one La' waiting on parts from BL)




Progress to date. One thing about this kind of project, it seems to never end.

Regards,

Doc

[nxtboyiii]

Maybe you could figure out a simpler way to build a biped...
BTW, you are very good at that!

[doc222]

Maybe you could figure out a simpler way to build a biped...
BTW, you are very good at that!

There are simpler bipeds. "Make things as simple as possible, but not simpler" , was said to be a quote from A Einstein.

Few changes to the pelvic/hips, skeletonized the studed section using transitional technic parts.
This also lowered it by one brick and shaved some weight off. The older versions had trouble pulling
this off do too weight shifting issues and weight of the upper section.

sumarsault