How To Build FAST Hexapod Robots! Fast Walking DIY Spider Robot
Вставка
- Опубліковано 24 лип 2024
- In this video I tell you how to use faster servos to build a faster hexapod robot.
My Discord:
/ discord
My Patreon:
/ makeyourpet
My Website:
www.makeyourpet.com/
0:00 Intro
0:24 Fast servos
0:40 Servo spec comparison
1:14 Update the config
2:06 Downside of fast servos
2:18 Overshoot
3:01 Bad impact absorbance
3:44 Switching back to old servos - Наука та технологія
That spider has the jitters. Needs less coffee ☕️ 😅
Thanks mate. I was going to buy those faster servos. Now, I need to think it twice.🤔
Excellent job.
This is so cool!
You may be able to correct for the jittering in software if you had analogue feedback on the servos, the jitters should show up in the feedback and allow you to dampen them somehow in the software. You may also be able to add some software compliance too based on the position feedback. It wouldn’t be as good as hardware compliance or current/torque controlled actuators though.
yeah that would be great and if the jitter shows up in the feedback I could probably do something about it in the code. but the downside of that is I will need 18 analog inputs for the 18 servos...
Is analogue feedback servo more expensive?
U r really good job 👍
You can redesign the phone holder to add a mirror reflection underneath the phone camera, and update the app to stream camera view
Yes, that exact design is coming to the channel soon.
Hello! I've just found your channel and brainchild, I love it! Having just seen this as second video so far, may I ask why not use something like a drone motor, if they're equivalent to cyber robotics? I love the cheapness and ease of the build, and by ease, I mean finding the resources. It looks a great project! Thanks so much for sharing
Thank you! to achieve the required torque this project is using servo motors, which is basically a motor + gearbox. Drone motors do not have a gearbox.
I was really wondering about how fast I could make it but you red my mind we must have similar interests
I think it could be even faster than what I show in this video. I am not even using the fastest servos out there.
@@MakeYourPet true but cost is a big factor
I try this hexapod some reason i can't get download program into 2040 servo board. just paper weight now. I gave up.
hey quick question can i use the servos MG996R instead
Hmm, there’s a lot wrong with the way most hexapod robots hold themselves too. Ppl always go for (arachnid) leg anatomy, but that requires a high base position for good agility, spiders achieve this by having passively constricted leg joints so they spend no energy standing up on all limbs.
But (insects) specifically, don’t have radial symmetry, which is another big reason these are usually slow. Hopefully those higher end motors you mentioned ever get implemented into properly bilateral hexapod designs in the future by someone, cause that’d be the key to a truly speedy bug bot without it being absolutely tiny.
The problem with overshoot is a PID issue with the servo. The problem is that youll have to replace the internal drive controller to mess with the motors pids. Totally possible though just a lot extra work
True. My guess is that these RC servos are tuned to function under load, like in a RC car. That's why I have the most issue on the coxa servos where there is no lateral load on the servo and it oscillates as a result. But it seems the faster the servo, the more obvious this issue it.
@@MakeYourPet Does that mean that you could use faster servos for all except the coaxial one?
It would be so easy to just adjust the gains. Too bad they won't let you do that easily.
I’m more than happy to pay for the STL files and design notes. Can it have more/less legs if desired? Where can I get ahold of the specs and manufacturing documents? I have money.
can you add your spidey a robotic arm so it can collect apple from tree.
I have a standard bluetooth game controller for phones (iPega PG-9023S) I use it often in my own robot projects. What I mean is to pair the joystick with the phone and control movement without touching to phone.
It shouldn't be difficult to include these types of joysticks in the software? Would you consider adding this feature?
Yes, that is in the plans. I was thinking to start with a PS4 controller.
No idea why this video was recommended to me, I understand next to nothing of robotics, but these bots are adorable. I'd want one as a pet if it had the actual AI to behave as one.
Happy you like it and welcome to my channel! You need no robotics knowledge to be able to build one of these.
you will even make it jump holy. getting scary now
You could eliminate those issues by buying brand servo made for rc cars. I'm thinking servos like Savox, they are fast and precise
Will this android control work on other hexapod frames. I don't have have access to a 3D printer. I do have a built hexapod. Adding switches to this robot will not be difficult.
Yes it should work. You just need to modify the geometry values in the config to reflect the parts that you are using.
So DS3225 are quicker than DS3235 but WE may have troubles
We could use DS3235 PRO but the price is more expansive ? Such a Big difference ?
I haven't tried the pro versions so I can't give you a definite answer. But If I have to guess I would say the faster servos will have overshoot. But if you just want to zoom around and go fast overshoot doesn't matter as you see in the video
Do you think it would be possible to create a walking speed hexapod robot that could carry any significant load? keeping the same sort of design ethic.. electric, servo motors.. etc ??
I've been following your work for a while but only just subscribed as Im looking to start my small project soonish. thx :)
It all comes down to the servo torque to robot weight ratio. for any significant load, you will need much stronger servos/geared motors.
Please help me out with how to connect two PCA9685 with raspberry pi for a hexapod project?
Join my discord server for questions and help.
hey. umm so I have been watching a lot of your videos, but I'm always left clueless as to where to start, and what to even look for on the GitHub page because I don't really understand. I see people building it, but I still can't figure it out. Is there a way for me to get a detailed explanation on how to build it ? like what parts do I need , If I don't have a 3d printer can I still get the parts? how to implement the code ?,how its connected to the phone? a detailed explanation on where each part goes because I have only half a years' worth of experience coding, so I'm practically a lay man here . I soooo want to build this, and I really love the content. Please can you make it more understandable?
Thank you for the nice words. If you need more help and details, definitely join the Discord server and ask there. I and many others are usually online and available to help.
How much cost this hexapod ❓ what do you control ❓ with PS2 controller ❓ please let me know thanks have a good day IS VERY IMPRESSIVE
Join my Discord server for the answer to all of those questions.
What working time on one battery?
around 10 minutes.
What is the purpose of using tip switches?
To sense when the leg reaches the ground. Since the ground may not be flat.
Aw, it cold. Poor baby.
😆
Has anyone tried to run the Robot with a sonoff nspanel pro (android based switch ~$90) instead of a phone?
I don't think so. Hopefully a RPI port will come soon.
Gives for the build a doku.
2:24 Please calm down bro
I love your hexa but especially firmware! I want to redesign it but with your firmware. I need the lengths from joint to joint for your already calculated kinematics codes. I measured coxa, femur, tibia's lengths from STL files it is 43, 80, 130mm respectively. Can you confirm this?
Thank you. Yes those are about right. You can find the exact lengths from the config file for each robot on my github page. You should be able to use any size leg parts and enter the values in the config.
IMHO you are missing the great opportunity to use a simple neural network to compensate for legs weight and backlash forces by letting the AI to predict the real force curve needed to drive the servos in place. If you use only raw data mathematically you'll never achieve good results and a real life "biological" adaptability to different surfaces/terrain.