/robowaifu/ - DIY Robot Wives

Advancing robotics to a point where anime catgrill meidos in tiny miniskirts are a reality

JulayWorld Attention-Hungry Games™ #1

Nominations are now open!

Help Julay.world stay up.

JulayWorld onion service: bhlnasxdkbaoxf4gtpbhavref7l2j3bwooes77hqcacxztkindztzrad.onion

Max message length: 32768

Drag files to upload or
click here to select them

Maximum 5 files / Maximum size: 20.00 MB

no cookies?

(used to delete files and postings)

Bipedal Robot Locomotion General Robowaifu Technician 09/15/2019 (Sun) 05:57:42 No.237
We need to talk about bipedal locomotion. It's a complicated topic but one that has to be solved if we are ever to have satisfyingly believable robowaifus. There has surely already been a lot of research done on this topic, and we need to start digging and find the info that's out there. There are some projects that have at least partial robolegs solutions working, but none that I know of that look very realistic yet. We likely won't come up with some master-stroke of genius and solve everyone's problems here on /robowaifu/, but we should at least take a whack at it who knows? We certainly can't accomplish anything if we don't try.

I personally believe we should be keeping the weight out of the extremities – including the legs – while other anons think that we should add weight to the feet for balance. What's you're ideas anon? How do we control the gait? How do we adjust for different conditions? What if our robowaifu is carrying things? What about the legs during sex? Should we focus on the maths behind MIP (Mobile Inverted Pendulum), or is there a different approach that would be more straightforward? A mixture? Maybe we can even do weird stuff like reverse-knee legs that so many animals have. Robofaun waifu anyone? What about having something like heelys or bigger wheels in the feet as well?

I'm pretty sure if we just put our heads together and don't stop trying, we'll eventually arrive at least one good general solution to the problem of creating bipedal robot legs.

ITT post good robowaifu legs

>tech diagrams sauce
Open file (14.92 MB 800x600 Healthy_2D.mp4)
Open file (90.53 KB 352x768 nms-simulator-2.png)
Neuromechanical simulation of human locomotion (SimGait + HBP)

Open file (415.45 KB 3503x2478 snapshot_V2.png)
Multi-Modal Locomotion
By the same lab. Includes executable simulation software links (Win, Mac).

Related papers:

Walking controller
>Imprecise dynamic walking with time-projection control

Human-like animations
>Scalable closed-form trajectories for periodic and non-periodic
human-like walking

Neural Control of Balance During Walking
>Neural control of standing balance has been extensively studied. However, most falls occur during walking rather than standing, and findings from standing balance research do not necessarily carry over to walking. This is primarily due to the constraints of the gait cycle: Body configuration changes dramatically over the gait cycle, necessitating different responses as this configuration changes. Notably, certain responses can only be initiated at specific points in the gait cycle, leading to onset times ranging from 350 to 600 ms, much longer than what is observed during standing (50–200 ms). Here, we investigated the neural control of upright balance during walking. Specifically, how the brain transforms sensory information related to upright balance into corrective motor responses. We used visual disturbances of 20 healthy young subjects walking in a virtual reality cave to induce the perception of a fall to the side and analyzed the muscular responses, changes in ground reaction forces and body kinematics. Our results showed changes in swing leg foot placement and stance leg ankle roll that accelerate the body in the direction opposite of the visually induced fall stimulus, consistent with previous results. Surprisingly, ankle musculature activity changed rapidly in response to the stimulus, suggesting the presence of a direct reflexive pathway from the visual system to the spinal cord, similar to the vestibulospinal pathway. We also observed systematic modulation of the ankle push-off, indicating the discovery of a previously unobserved balance mechanism. Such modulation has implications not only for balance but plays a role in modulation of step width and length as well as cadence. These results indicated a temporally-coordinated series of balance responses over the gait cycle that insures flexible control of upright balance during walking.

<Keywords: balance, walking, neural feedback, vision, virtual reality, sensorimotor control

Open file (126.60 KB 560x477 zad0021028820002.jpg)
Dynamic Principles of Gait and Their Clinical Implications
>A healthy gait pattern depends on an array of biomechanical features, orchestrated by the central nervous system for economy and stability. Injuries and other pathologies can alter these features and result in substantial gait deficits, often with detrimental consequences for energy expenditure and balance. An understanding of the role of biomechanics in the generation of healthy gait, therefore, can provide insight into these deficits. This article examines the basic principles of gait from the standpoint of dynamic walking, an approach that combines an inverted pendulum model of the stance leg with a pendulum model of the swing leg and its impact with the ground. The heel-strike at the end of each step has dynamic effects that can contribute to a periodic gait and its passive stability. Biomechanics, therefore, can account for much of the gait pattern, with additional motor inputs that are important for improving economy and stability. The dynamic walking approach can predict the consequences of disruptions to normal biomechanics, and the associated observations can help explain some aspects of impaired gait. This article reviews the basic principles of dynamic walking and the associated experimental evidence for healthy gait and then considers how the principles may be applied to clinical gait pathologies.

Open file (580.75 KB paper.pdf)
Modeling robot geometries like molecules, application to fast multi-contact posture planning for humanoids
>Traditional joint-space models used to describe equations of motion for humanoid robots offer nice properties linked directly to the way these robots are built. However, from a computational point of view and convergence properties, these models are not the fastest when used in planning optimizations. In this paper, inspired by Cartesian coordinates used to model molecular structures, we propose a new modeling technique for humanoid robots. We represent robot segments by vectors and derive equations of motion for the full body. Using this methodology in a complex task of multi-contact posture planning with minimal joint torques, we set up optimization problems and analyze the performance. We demonstrate that compared to joint-space models that get trapped in local minima, the proposed vector-based model offers much faster computational speed and a suboptimal but unique final solution. The underlying principle lies in reducing the nonlinearity and exploiting the sparsity in the problem structure. Apart from the specific case study of posture optimization, these principles can make the proposed technique a promising candidate for many other optimization-based complex tasks in robotics.

<Terms: Humanoid Robots, Kinematics, Biologically-Inspired Robots, Task Planning, Gesture, Posture and Facial Expressions, Optimization and Optimal Control

Edited last time by Chobitsu on 11/28/2019 (Thu) 16:54:52.
I think the most fundamental thing for walking is balance. It could be optimized for as the minimum amount of energy to maintain a pose or state. It's really important we acknowledge robowaifus will have different bodies and parts than us and what they find balanced to walk with will be something else entirely unless we pay close attention to engineer their bodies with similar weight, range of motion and forces as us.

The second most important thing is experience. Robowaifus will need to practice to improve their world model's prediction function to judge the weight of objects, their weight distribution, slipperiness of surfaces and such to understand what the minimum energy state will be. There are things robots don't have sensors for like feeling a cylinder sliding through their grip, but there are visual cues and other things that can be detected such as slipping objects feeling lighter. Robots can also have completely different sensors in their body parts such as accelerometers and gyroscopes.

Objects in general need to be approached with caution and gentleness, using only the minimum amount of force needed to achieve tasks. Overexertion requires corrections and can be identified within the system as shaking. In disaster situations rubble and debris may also move when walked upon. Motion planning will need to navigate uncertain environments and avoid potential motor stalls and falls by taking the safest understood route, unless there is no other way to go or it's too dangerous and should stop. Collisions with other people, stairs and moving vehicles must be also factored in but I think finding the minimum energy required will naturally solve this because walking into a bus going 50 kph is gonna be a lot of energy to overcome.

I wish I had some experience with 3D programming. If someone could put together a simple mechanical simulation or find one, I could rig up some AI to first master using its limbs and then optimize it to walk towards a goal with minimum energy. It would help us study how limb weight, weight distribution, joint placement, actuator power and range of motion will affect gait and we can come up with better designs that can balance more like a human being or something else entirely that's beautiful and practical in its own way.
>If someone could put together a simple mechanical simulation
<or find one,
would this one do?
Open file (33.52 KB 246x251 simulation.png)
Not sure, I'm on Linux and would have to see if it can run in Octave. I do most of my AI in PyTorch and it seems Octave has a Python interface.

I think this will be helpful but I was thinking more of a simulation where weights can be attached to body parts for more realistic torque calculations. That way we can test, design and evolve parts to be more functional in the real world.
Well, I'm setting up Octave myself for the very first time right at this moment to try and test the program, so you're ahead of me heh.

Yes, that diagram helps. I'm think about building a 3D model for robowaifus starting with a graph description of the bones for rigging using boost.graph. Once that's workable to the degree where mesh data can be bound to the bones and rendered, do you think you could work with it?
Here's the error I'm getting trying to run the m script w/ Octave:
johnny@mactoshub:~/_msc/rw/repos/Walking3LP$ octave Walking3LP.m
error: 'gui_mainfcn' undefined near line 62 column 5
error: called from
Walking3LP at line 62 column 5
BTW, if you examine the GUI image, you see that mass can be added/removed from limbs for the sim runs. Looks very flexible to me tbh.
Here's the code section in question. The error line mentioned is the one inside the else clause. Apparently 'gui_manfcn()' is a Matlab API which Octave must have a different name for. Any matlab guys here?
% Begin initialization code - DO NOT EDIT
gui_Singleton = 1;
gui_State = struct('gui_Name', mfilename, ...
'gui_Singleton', gui_Singleton, ...
'gui_OpeningFcn', @Walking3LP_OpeningFcn, ...
'gui_OutputFcn', @Walking3LP_OutputFcn, ...
'gui_LayoutFcn', [] , ...
'gui_Callback', []);
if nargin && ischar(varargin{1})
gui_State.gui_Callback = str2func(varargin{1});

if nargout
[varargout{1:nargout}] = gui_mainfcn(gui_State, varargin{:});
gui_mainfcn(gui_State, varargin{:});
% End initialization code - DO NOT EDIT
>gui_mainfcn() *
Open file (603.31 KB 866x523 quattroped.png)
There's a simple but inelegant short term solution; add an extra limb with omnidirectional wheels to the bipedal robot. It can be an armature that's detachable at the hip with the wheels behind the body. Or it can retract by folding inside the lower or upper legs depending on the kinematics.

This would allow for decent mobility at a slightly faster than shuffling pace as 3 points of contact on the ground at all times creates static rather than dynamic locomotion. It also lets you add a lot more weight to the robot without worrying about stability.

This is the only robot I've seen that tries a hybrid folding limb approach and is pretty innovative. As for a simple mechanical simulation program I'm also trying to find something that does this. playdynamo.itch.io/dynamo looked promising but it performs terribly in Wine and keeps trying to connect to the internet when blocked. I might give FreeCAD or Blender another try but both of those are very time consuming to get used to.
clever idea bout the wrap-around wheels heh. we've had numerous discussions on the alternatives to bipedalism for our robowaifus here over the years. i think the general consensus for those of us not willing to give up on a human-like form for our waifus like myself :^) the nearest workable alternative seems to be to give her a motorized wheelchair to move around with until we can perfect bipedal locomotion.

Still, good ideas anon keep them coming.
BTW, I think the Pepper robot uses a tri-wheel scheme for it's base?
I see my approach more as an assisted form of bipedalism building up to the real thing. This quick sketch might give a better idea of what I'm talking about. It would still have the appearance of walking so it's not comparable to sitting in a wheelchair or being carted around vertically on a hand truck.

By using two mechanum wheels you get to spin around by running each wheel in opposite directions. Omnidirectional wheels have their own advantages and disadvantages.
70.000 USD for a fucking prosthetic leg, jfc.
I see. Yes that diagram does make it clear. In effect, you are providing her with an irl cartoonishly oversized 'foot' with which to get around on, but one that would shrink back to normal size when not needed.

Ever price even the most basic of medical supplies in a hospital context anon? No surprises tbh.
>Ever price even the most basic of medical supplies in a hospital context anon? No surprises tbh.

Yep, but 70k is still a lot for a device that's not even worth a fraction of the price in materials or manufacturing. I mean what's to be expected when every megalomaniac cocksucker and their granny is allowed to milk people's health for all it's worth and call it a day.
Look I'm not promoting their (((approach))) I'm simply encouraging you to be realistic. Let's continue this off-topic discussion in the Lounge if you'd like.
Open file (831.34 KB 755x902 softfoot.png)
Open file (14.98 KB 173x509 EMIEW2.jpg)
Open file (237.35 KB 373x624 WL-16.png)
Looking into humanoid bipedal robots that have a single wheel on its feet this EMIEW2 is the closest one. Other robots typically have several wheels per foot(Zephyr) have the wheel as part of a normal robot foot to add wheeled locomotion(GoRoBot) or more than a pair of legs. The other decent design is the WL-16 leg platform using WS-2 feet but that has a caster alongside a powered wheel.

I've gone through at least 50 research papers on robot walking methods(and have another 50 to go through) and have seen some interesting designs and solutions but none of them use the approach I'm proposing where the wheel is only used to shift the center of gravity during the stride while keeping the robot stable with 3 points of contact on the ground at all times.

One thing that hit me when browsing another forum are those minisegways cost less than $200 now. Rather than trying to do bipedal locomotion I'm going to have my first model balance properly on top of that until a cheap solution is found.
>I've gone through at least 50 research papers on robot walking methods(and have another 50 to go through)
Mind posting a few of them here for the rest of us anon?
Yep, we've had this conversation before now. The posts haven't been restored here yet but I came down in favor of the idea myself.
Why is that robot in the middle wearing a diaper?
Eh most of them aren't worth reading otherwise I'd have uploaded a pack of them here. These two are probably the best I've found on the topic for a general overview.

The segway mini isn't really a hoverboard since you use your knees rather than your feet to control it. The center of gravity is much lower to the ground and there's no having to keep the robot upright on it.

It looks even more like a diaper from the back. The reason for the protrusion around the abdomen is it's designed to sit in the seiza position with giant feet.
>Eh most of them aren't worth reading otherwise I'd have uploaded a pack of them here. These two are probably the best I've found on the topic for a general overview.
Fair enough thanks, thanks for filtering the mediocre for us Anon.
>segway mini
They seem a little more expensive than 'less than $200'. Am I missing something? They do look like a good choice. The robowaifu would still need basic bipedal locomotion to be able to use it effectively (climbing on/off, steering) but nothing particularly good. Good idea.
Open file (18.97 KB 248x500 Roll.jpg)
A lot of good thinking going on here anons, but much of this discussion is fundamentally flawed. We want our waifu to be: -Inexpensive -Easily Reproducible -Low Power From these tenets, having any active stabilization beyond the minimum isn't advisable. She should have legs that naturally conform to walking on various terrains. For this reason, having a tail or some other method of passive balance (a dress that hides balancing struts, big feet with a low center of gravity, or have her hold something like a walker in front of her.) these examples will all work for a robowaifu that is inexpensive and low power as she doesn't waste energy on balance. Roll is a great example of low center of mass with big feet
>>1873 I don't think balance is a concern for complexity or power consumption. Returning to balance from outside forces is easily accomplished. Getting a dynamic gait is the big problem. Maybe we could cheat it with a neural network. Really if you're not just making a sex robot you're going to be using neural networks or something close. A simple solution for movement is to go completely on rails. Not rails literally but railings countertops and other waist-height edges. The waifu would be restricted to the home but would have plenty of well defined static supports to use and only flat surfaces to traverse. She'd just never cross a wide open room or maybe you could throw a pool table in the middle of it. Some Anon have entertained the idea of using a tether for power and could equip their home with a literal overhead rail system that would be far more discrete than the other stabilization ideas mentioned.
>>1882 If you're using rails and tethers, why walk? Also, self balancing is easy for humans but, hard for computers. Neural networks should be used for her personality and communication, it's better to make her walking something that relies on good mechanics rather then complex software. We must keep costs and complexity both in hardware and software, to a minimum.


Captcha (required for reports and bans by board staff)

no cookies?