Rawbots Club Forum

Rawbots General => General Discussion => Topic started by: Doctor Virus on March 12, 2016, 02:44:25 pm

Title: Walking bot group project
Post by: Doctor Virus on March 12, 2016, 02:44:25 pm
Hello all.
Today I had an idea about rawbots and thought to myself, "You dont see much walkers.." so I have a group project idea if anyone would like to work on something with me to make a walking bot that doesnt slowly inch forward, a bot that can walk normally around the planets.

I hope I put this in the right forum section
Title: Re: Walking bot group project
Post by: MarvinMan on March 12, 2016, 03:47:03 pm
That sounds like an interesting idea. Making something walk at all should be fairly simple, but there's definitely a lot of work to do to make something that's smooth and controllable.

What sort of walker were you thinking of- bipedal or hook footed spider-bot? I think a lightweight bot with 6 legs is probably a good starting point.

The only biped I remember seeing had to use hovers to support the body. I built a huge spider bot before, but I lost the save before I got a chance to let it loose on one of the stardust planets.
Title: Re: Walking bot group project
Post by: z26 on March 12, 2016, 05:16:24 pm
It depends.  What do you consider fast?


Personally I consider this reasonably fast as far as walking bots are concerned (except uphill, maybe using the small wheels as feet would help with traction).  Certainly moves in a much more efficient, lifelike manner than that mindstorms nxt bot I've made once.  I actually think walkers in rawbots are relatively easier to build than in other sandbox games, the engine is suited for it.  But that's just my opinion.

Also, if you want to do a bipedial walker, shifting a mass around for balance might be more elegant than using hovers, but thats also subjective.  an ostrich body shape may work better than a human one?  If you do a walking bot that has to support some weight, you should use motor based hinges since they are a lot beefier than elbows are.  Heavy continuums also offers strong joints, so they might help if you want something sturdy.
Title: Re: Walking bot group project
Post by: Doctor Virus on March 12, 2016, 05:28:52 pm
Ah yes I remember the spider bot. Though my idea is to make a car bot that can run, (tribute to my special car lover) though we all know balance is one issue along with making a repeating system run safely and steady

This could be a big move foward to make things with fully functional legs I believe
Title: Re: Walking bot group project
Post by: z26 on March 12, 2016, 05:36:01 pm
Now I want to give a try at building a walker because of you.
Title: Re: Walking bot group project
Post by: MarvinMan on March 12, 2016, 05:39:11 pm
Wasn't that one of Trooks bots? That was the sort of thing I was thinking would work best, perhaps with slightly longer legs and more ground clearance.

Using wheels as feet would work for larger bots, but they seem a bit big and heavy to be swinging around on smaller bots. I think the trick to getting good traction and smooth movement is to make sure there isn't any dead time in the cycle in which the foot is dragging on the ground or going in the wrong direction.

I think building a bipedal bot should probably wait until we get a more stable platform walking well.
Title: Re: Walking bot group project
Post by: z26 on March 12, 2016, 05:51:58 pm
as far as multiple legs walking there are two stances we possibly can take.  Crocodile style sideways legs and inline legs like mammals have.
crocodile legs are more stable while inline legs are faster.  with that said there are other factors influencing that tradeoff like body width leg length etc.  a narrow milliped bot is better with a wide stance but a wider squarer bot can do well with a narrower stance.
Title: Re: Walking bot group project
Post by: MarvinMan on March 12, 2016, 05:53:43 pm
Turns out I didn't completely lose the spider tank. Unfortunately, it seems that the version I found has all of its legs trying to walk in different directions.

(http://i.imgur.com/RX7HLpP.png)
Title: Re: Walking bot group project
Post by: MarvinMan on March 12, 2016, 06:05:11 pm
Crocodile style walking with a flexible spine would probably work well for a small bot and could use fewer actuators, although I don't know how you'd steer it.

I'd be inclined to go with something closer to inline legs as you'd be able to have a rigid chassis and better control over the walk cycle should be possible.
Title: Re: Walking bot group project
Post by: z26 on March 12, 2016, 06:09:34 pm
Maybe steering could be done in a ackermann like fashion.

Also, I've just noticed how slow the default angle pid can be...  One problem is that in real life pid updates at a very fast frequency, but in rawbots you're limited at the physics engine timestep essentially.  Makes designing a fast AND accurate control system tricky.

That thing is pretty massive.
Title: Re: Walking bot group project
Post by: MarvinMan on March 12, 2016, 06:23:21 pm
Elbows on their own are pretty hopeless. Using a controller to set the velocity to hold an angle works well, but makes the code big and can become unstable.

I think using pistons to provide most of the propulsive force is a good thing to try and do, as they are very strong without a controller and don't tend to explode. Where possible, I usually try to use mechanical linkages or make things naturally stable so the physics engine does all the hard work instead of the VP code.
Title: Re: Walking bot group project
Post by: z26 on March 12, 2016, 06:30:18 pm
So basically it works like a real life excavator, using linear pistons to make the articulations rotate?  Thats a pretty cool design.
Title: Re: Walking bot group project
Post by: MarvinMan on March 12, 2016, 06:35:12 pm
Pretty much. The main disadvantage is that you still need the motors/elbows as hinges to let the structure move.

If we have piston legs, there's a good chance of being able to jump, although it may just end up ramming the feet through the terrain.
Title: Re: Walking bot group project
Post by: Doctor Virus on March 12, 2016, 06:56:34 pm
I can imagine the pain a bot builder would feel watching their creation get stuck in the ground. I move all this talk though and the styles in which these bots were built
Title: Re: Walking bot group project
Post by: z26 on March 12, 2016, 07:21:10 pm
One relatively easy way to cheat the physics for stability would be to weight down continuum feet when touching the ground and lighten them up when they're leaving the ground, but that just feels wrong.  I think that while the ability to change the weight of continuums is great by itself, the ability to do so with code in real time feels cheaty.  In my ideal world, you only should be able to update the weight of a continuum by hand (as for instance the type of a math operation)
Title: Re: Walking bot group project
Post by: MarvinMan on March 12, 2016, 08:54:33 pm
Things going through the walls/floor usually only happens with terrain hexes, which can be deleted to free the bot. At one point I had a grenade cannon that could reliably fire straight through walls.

Varying the mass of continuums on moving limbs could get messy, especially when combined with the amount of force that pistons are able to provide. You can actually make very effective engines with just a continuum on the end of a reciprocating piston. I agree that it seems like cheating to vary mass on the fly, and you may as well use hooks instead, which may let the bot walk on walls or the ceiling.
Title: Re: Walking bot group project
Post by: cupid_the_conqueror on March 13, 2016, 01:50:45 am
If your worrying about balance, remember that one of the odd things about unity physics is that it LOVES the gyroscopic effect trying spinning a single contuim at 50rpm when it has 100 mass. It'll twist the crap out of whatever it is attached to!
Title: Re: Walking bot group project
Post by: MarvinMan on March 13, 2016, 09:07:39 pm
I don't think rawbots still uses the default unity physics. IIRC, it got replaced with the bullet physics engine.

I've not noticed anything gyroscopic happening with spinning masses before, but it could be interesting if it does work.

I built a leg on a test platform, and came up with a few observations:
>The foot must have matched the bots speed before it touches the ground, and must be lifted before the leg finishes the power stroke.
>Grenades make poor feet, and continuums work ok but aren't round.
>Small wheels make good feet, but are quite big
>The fast leg movements should use an open-loop controller, with a slow closed-loop system to tune the settings of the fast loop
>Synchronising multiple legs could be tricky, especially when reversing/steering.
Title: Re: Walking bot group project
Post by: Doctor Virus on March 13, 2016, 10:41:43 pm
Ah yes steering is one big issue but some pistons and hooks should do