I really hope the phone continues to get bigger lmao

This is really impressive for just a couple weeks. It's got some quite sassy poses, too.

Pv case has gotta be the most random small marketed ad ive seen on KZfaq 😂

The switches-in-the-feet idea reminds me of how velocity is determined in musical keyboards. Each key has a pair of contacts that make connection one after the other, and the timing between the two correlates to how fast the key is pressed. Might be worth looking into.

.. Okay that phone gag is getting larger then life and I am here for that!

"My robot does really weird things" Sounds like the opening scene of a dystopian SciFi horror movie.

The way the head moves reminds me just a little bit of those videos of chickens keeping their head in place while the body moves. Actually the whole build is somewhat chickenesque, in size and shape. Pretty cool!

You need to make a droid that follows you around holding the phone for you!

James: "I'm not too unhappy with this for two weeks work." Me: Wondering if I could build even just leg and if the tech i used to build it would even be relevant by the time I finished.

I think the waddling/shuffling motion would work really well for a GONK!

something about the way it moves is just adorable, he's even cuter than the originals.

Noticing that your table is wobbly, trying it on the floor to see if it works better on a stable surface, and then finding out that it only works on *wobbly* surfaces is hilarious XD

Brb waiting for the video where he needs a crane to hold his phone

This is fantastic work. Well done, James! 👏🏻

These robots you make are awe inspiring James. I can't wait to try something like this myself with the kids.

This is funny because I said to myself at the beginning of the video that you should probably program and test it on the floor rather than the table. And turns out at the end it only worked on the table and not on the floor. I had a little laugh.

Amazing project as always.

Loving the giant phone James! Really cracked me up 😂😂

Thank you for making this series! You inspired me to try and make a walking robot similar to yours for my robotics course.

great video as always james!

I always thought it would be neat to make a 6-DOF "arm" that can crawl like an inch worm and collaborate with others to do complex tasks. Something like the "microbots" in Big Hero 6. You can imagine them having interchangeable actuator mounts on either end so they can switch out graspers and other tools. You could also imagine six of them connecting to a tool box or something and walking it around like legs. My dream would be to have these things be able to build an unbounded volume with both additive and subtractive manufacturing. I'm sure the software for choreographing such a swarm would be crazy complex.

oh, this rocks! best project yet, nailing it 😎

Thats awesome! I really liked the Star Wars droids/Robots - its really cool to see an open source version of them :)

Simply amazing... I love your videos! Keep on going!

Strict adherence to linearity and angularity. Between the bearings and motors there is plenty of inspiration toward circular, spherical and other non-linear opportunities, but this feels more like adding gears and axes of movement that ultimately just reinforce a cubic vertical/horizontal-only palate. wobbling back and forth as a kind of cute side-effect that can approach a form of walking is what we get.

At @10:30 I was like, maybe you should try a PID-controller and BAM! Your next sentence was "I'll use a PID-controller..." Love your videos and how you gradually solves problems throughout your projects, making them more and more complex!

Excellent video, great project - I'd honestly love to see what you could manage making this style of robot using your cycloidal drives, or at least something more tune-able. It might not be the most practical - but it's a very characterful robot, and I can see situations where it could even be useful.

I used your Astromech project to start my R2-D2 however many years ago that was. Might be time to build this lil guy next.

impressive design in this timescale Thanks for sharing your expirences withe all of us 🙂

I just wanted to say Thank you James for introducing me to 3d printing and showing me that building an R2D2 was possible to build in this manner. As you can imagine it's taken a long time to learn and then discover Mr Baddeley. But it's you who sparked off my long journey and I'm finally nearly there after 4 printers and several failed attempts. And I've always wanted to thank you for helping me to finally bring my dream into reality.

You're going to get hired as an Imagineer soon

This is so cool !!! It's like a super power to spawn an army of adorable machines. You're awesome Sir.

Great work! Big fan of your channel. I would recommend you use a Kalman filter or a complementary filter to fuse IMU data.

Yes mate this build is exciting. I'd say you can run with this mission for a long long time. Two thumbs up!

That droid looks super cool, and I can only imagine the code and math behind controlling all that! You've made quite the land walking robot, but with the design of the feet, it feels like it shouldn't be too hard to retrofit them with skate blades and get it to move around on ice! Especially with the new leg servos, I think it has all of the necessary movements to balance properly!

Really looking forward your take on making a smaller cycloidal drive

You microphone went up to a new level of ridiculousness! Love it!

This is awesome work. Have you thought about switching the head movement axis to side-to-side so you can damp the rocking with head movement? Similar to TIDs used in skyscrapers?

Hey! Love the content, have you ever thought about making a car suspension with the cycloidal drives and tune the motors to absorb more load in off road mode and less in sport mode?

As soon as I saw them showing at the Disney parks I knew you would be building them! Love it james!

I'd love to see this project continue, and overcome the instability problems. I thought one of the advantages of Dynamixels was that they could be backdriven and implement a 'virtual spring'. I also thought the user had some control of the PID parameters, enabling the 'rigidity' and damping to be varied. I've never used them, so I could be mistaken. There is also a possibility to build mechanical springiness and damping into the leg structure.

If you want to eliminate gear backlash, copy what's used in the head drive of many optical drives. One of the gears is two thing gears in a stack, with a spring that causes them to rotate slightly in opposite directions. Even if the spring collapses during high speed movement, when the head slows to tracking speed, the spring spreads the teeth to eliminate backlash. A robot foot idea I've had kicking around for ages is a passive locking design for rough terrain. Without weight on the foot, springs hold it perpendicular to the lower leg. As the leg sets the foot down, the foot passively tilts in 2 axes to match the surface angle. As the full weight of of the robot presses down, the tilt angles of the foot lock. Just above the 2 axis tilt would be a fore-aft joint with springs just stiff enough to allow the bot's momentum to tilt forward over the foot. That joint would also have a clutch to control the ankle speed or to lock in place to keep the bot upright when it's not moving. Best stability would be with one leg slightly forward and one slightly back. The foot tilt axes could have angle sensors to feed information to the walking control. I figure that robot walking should work from the foot upwards. Put a leg forward and let the foot come down then take data from the foot to adjust the knee and the hip. What I've seen in most robots is a "dumb" foot the robot stabs at where it calculates the ground should be then it reacts to various sensors telling its controls which way it's falling over. A quadrupedal robot can get by with four feet it can treat as simple single points of contact. A bipedal robot with flat feet can handle flat surfaces, or uneven surfaces that are a bunch of flat surfaces of various heights. What I've seen of bipedal robots on uneven terrain that's made of angles or curves is they don't stop and stand on such surfaces. They either keep moving forward or "dance" by constantly pumping their legs. Taking input from foot angle sensors and having ankles that can both tilt in different axes and lock in place would enable steadier uneven surface walking and stable, motionless, standing. In my much younger years I lived on a 2 acre property with areas that were far from smooth. I used to run across them in the dark and never fell or twisted an ankle. I managed that by keeping my ankles loose, letting my heels land first then allowing my feet to flop down and conform their tilt to the ground. One I sensed my forward foot was planted I'd tighten up the ankle and allow my body weight and momentum to lever forwards while I picked up my rearward foot to swing that leg forward to repeat the cycle. I was quite the logical thinking teenager. :) Another thing I did was ditch running. About half the property was a grassy hill slope with a flood irrigation ditch along the top. the sides of the ditch were at about a 45 degree angle. I came up with a technique where I'd run with left foot on the right slope of the ditch and right foot on the left slope. I'd pull my rearwards foot outwards just enough to clear the other leg then swing it forward then push back across in front of the other leg. For me it was a fun and quick way to move along the ditch, almost self guided like a train track. Why did I do that? Because it was fun, faster than walking in the bottom of the ditch putting one foot in front of the other. I was also *that kid* who in PE class when it came to standing on one foot could just stand on one foot while the rest of the class was flailing their arms and free leg about, trying not to fall over. I was the shortest and lightest kid who could just manage a 12 minute mile run and was too light to compress the springboard even a micro-inch. How light? I could stand on top of a perfect, empty, 12oz soda can without crushing it. For fun I'd carefully stand on a can then give it a gentle tap with my other foot to make it collapse.

You could try mounting the gears on springs. Some side to side inertial dampening with the head would probably be good too. Cool project. 👍

Pretty neat! Thank for the excellent work!!

I really love this robot - fantastic work! Disney really struck something very loveable with those droids. I would love to see you do another spin of this design with the brushless/compliant joints, then spend another episode diving into the kinematics and puppeteering of the droid. Maybe even a fourth episode adding some of your earlier Jetson AI vision for some responsiveness!

Hello from New Zealand James! Your channel has me captivated every new thing you build! Your cheap hack worked perfectly for the balance and made me want to ask if you considered having either something in the head moving on the other axis or maybe adding tilt axis to the head to help in the catching of the sideways issue? Anyways carry on your awesome work can't wait till the next vid!! :)

So awesome , love it

My favorite engineering channel

Awesome build. Any updates on the Open Dog project?

Looking great overall! & I'm sure nature does all sorts of dirty hacks too ^^

Excellent work!!!! 🎉🎉🎉🎉🎉🎉❤

Thanks for sharing.

Wow... you made it in couple of weeks along with filming and editing video , when for disney, it took years of research with rather high budget and a big team of engineers... This is on another level of impressive... also i really wish you would try building open dog 4... something that is comparable to spot....

You really need a larger, wobbly table

Great video. Amazing how you can get really close to gettting something working, but that last little 5% can be a right PITA. Maybe get Robotis/Dynamixel to take on the cycloid drives as a product!

So cool!

Nice build so far! Question I’ve had: What extruder do you use for your larger nozzle prints like 1.2mm? Your useful ROS robot videos got me investigating larger nozzles than 0.4mm but I doubt my stock tronxy setup can do the flow 1.2 would need.

Could you mount the IMU into some soft mounts to filter out the vibrations mechanically? And what about doing FFT analysis on the noise and seeing if there are any frequencies you could apply a notch filter to?

are you planning on using Apple Vision for designing?

There are only a few people I trust with robots. You are certainly one of them.

great video

Can't wait for the "I built a robot to hold my phone because it got too big" video series...

Learn sooo much from your videos Dude! Thank you! If only the world was run by engineers and scientist! 🙂

Very cool.

Hello, I really like seeing these projects. I have a question: What was your area of study? or what kind of courses would you recommend for me to take to be able to do this type of project? mainly speaking of the math and programing. Thank you for your time.

Could you run a loop of surgical rubber (latex tubing) as a shock absorber instead? - Have it act as a muscle tendon instead of a back driveable solution? (not just at the point where there is a problem I mean a loop that runs the length of the leg so thigh to ankle)

if you give it wings it'd be a robot chicken, and I can't unsee it.

10:25 got me laughing my ass out because of that bots posture.😂

as an idea, try putting the IMU in the head and then attempting to keep the head stationary. This is (roughly) how humans keep balance via the fluids in your inner ear and, if I had to guess, it allows you to easily and accurately account for small micro-vibrations/misalignments and essentially acts as a high-pass filter for physical noise. In other words, if you constantly try to keep the head stationary using *_just_* the neck motors then if you can't keep it stationary you know you need to use the leg motors to account for the remainder of the inaccuracy. Actually programming this would be a bit tricky and the naive approach would run your leg adjustments at a decent latency as you try to look at the past to try to figure out what portion of the readings are new head-wobble or old body-wobble, but hypothetically it could work.

Can you make torso keep straight vertically, should help balance.

Have you seen the video with the gyroscope balanced bipedal robot?

Disney really should hire you, because what took Disney several months and lots of top of the line engineers and unlimited budget to build, you made that alone in just a couple of weeks with limited resources !

Do you think you could do a smaller version ? I would love to print it and do it too!!

Could the feet be redesigned to be the damping? Make them like the bionic legs that were used to (unofficially) break running records?

I know Disney have all the money, but I'd love to see you continue with this and eventually recreate in your loft what they managed with all their resources.

I am very impatient to see when you going to make an AT-ST !!!!

it looks so cute crouched! though I must say I think it being crouched would allow for more than just looks; I don't know a lot so of course I'm trusting you know better but not only would full steps in an arc be theoretically possible if the range was used from there, but the lower center of balance might make it more stable. In the end I just think it looks cute with bent legs though :P

love the comment that it is alive

Awesome.

Could you use a gyro sensor instead of trying to use switch timing to back out the angle

Dynamixels should have a force compliance modes that does what you're asking.... have you looked at hte protocols supported? Are they willing to support you in setting the motors to be more compliant?

I've been following your channel for years. Glade to see you building robots again. One little suggestion which I asked before, you may try to bring your future robots to a next level by making them able to walk step by step slowly instead of rocking side by side. I know it's much harder, but it may be a good project to develop. Cheers.

James, hijacking this new video to get a proposal through to you: How about a vaccuum cleaner that follows you around, so you don’t have to tug on the hose all the time. It should be omnidirectional (without the wheels collecting too much dust) and avoid obstacles. Seems like just up your alley and would be extremely useful!

What is the trigonometry equations and how would I translate them into working with 3d vectors?

Can you make a mash up of all the time you take in cad I wanna do projects like t this and kinda do them but I feel I’m only using surface level tools and not being very professional so I would love to see just a long even if it’s Timelapse videos of you in a software so I can lean some good habits I might be missing or see where I can improve on what you do

Is there a reason to use back driving over suspension?

Within a few videos that red phone will get so big it wouldn’t even fit in the house anymore

James. We want a proper dinobot and we want it now. You are the alpha dog mate I trust in you Also cheers from italy

Would a swinging counterbalance help the stability of the robot? or are you trying to make it look very similar to the star wars units?

Every day am shuffling!

I'm wondering if a gyro would dampen out some of the slop

Maybe a Kalman filter could be useful for the IMU issue?

I don't know if you cover this, but in drones they tend to soft mount the IMU to try and dampen a heap of the same noise you're seeing here

What if you make it walk slower but longer steps. You could make it lean over to the opposite side of the leg it lifts and time it for when the feat hits the ground. Just like how a human does. Basically make it almost able to balance on one leg while the other is moved. Also the feet should be wider to minimise the sideways tipping. Then it would not rock back and fourth as much.

What is the best entry level 3d printer??

The moment it started and I saw the phone I thought "oh you cheeky bastard", also as always amazing work

Did you try retuning your PID controllers with the table on the ground?

Very impressive - although one day I’d like you to be able to train the robot to understand when it is being pushed about and complain (or even slap back!) 😀👍

Haha he built a table dancing robot that wiggles it bum, where do I put the the $5 tips 😂

At 12:00, for a "dirty hack" that does an awfully great job of dampening the oscillations. Honestly I think you're selling yourself short there, damping, and active damping isn't simple. I think switches might be too binary for ballance. You might get some crude jerky sense of ballance but for a dynamic sense of ballance I think like you discovered with opendog, back drivability and motor current reading is best.