Using Dynamixel prices out like 90% of your potential userbase

Hello,

I’m not a mechanical engineer. I’m a coder studying deep reinforcement learning.
Anyways, because reachy uses dynamixel, it’s prohibitively expensive.
If you guys could use some medium size brushless motors and ODrive’s,
you would probably get more people building reachy/ sections of reachy.

It really sucks when some tech project’s / toys are 1000+ dollars to build.
That prices out 99.9% of your potential userbase. Especially students.

If you drop the dynamixels you could cut the price in half.

Thanks for listening to my rant,
Gibson Martin

2 Likes

Hi,

Thanks for raising this legitimate concern and offering the opportunity for me to explain our position on this aspect.

We understand the frustration for everyone wanting to get their hand on a cool robot when the price goes far beyond the hobbyist budget. We’ve already experienced that with our previous humanoid project Poppy which was using 25 Dynamixels.

Maybe we will drop Dynamixels in the future but to be honest, it is not the problem.
Given the current state of robotics technologies, it is just not possible to offer more than a toy for less than $1000.

Personally I expect the “consumer humanoid robot market” to converge toward a cost between a high-end computer and a small car, simply because a robot involve complex hardware.
Inside, you will have at least a powerful computer, batteries, cameras, plenty of high performance motors, dozen of electronics boards, thousand of mechanical parts and so on… It is just not possible to get all of this for less than we pay our personal computer. At least if you want to respect workers building it and our environment in the process.

At Pollen Robotics we are aiming to offer convincing multipurpose robots for services and assistance. So to be totally transparent on our roadmap, we are not focused on cost-killing but rather on improving the state-of-the-art so we can do more with a $20k humanoid robot until it eventually becomes a reasonable investment for most of us.

But meanwhile how the whole potential userbase could be involved?

First, my comment concerns only adult-size humanoid robot but it is possible to offer more simple robot (yet lovely) for an accessible cost. We already did it and will probably do it again :slight_smile:

Second, while Reachy is not a reasonable investment for an individual, it is clearly a good one for a team in a fablab. You won’t gather people and won’t go far with a cheap yet shitty robot but you can with a high-end robot such as Reachy, and why not launch your own robotics company!

We are super happy to know more & more individuals are excited to add their energy in the collective effort to make robotics finally happen.
To those who want to do it right now, my message is : Don’t invest your money in low quality product but rather invest your time in the nearest fablab. Motivate people around you, launch a collaborative project and find local budget to buy a proper platform to work with (Reachy or another robot). Robotics is too complex and expensive to go alone, bring friends with you and you will really enjoy this exciting journey!

3 Likes

Hi, thanks for tanking the time. I can see the value in having great motors and ensuring the experience is optimal since day 1. I just learned about, reachy, I have a couple of 3d printers, I have many raspbery / orange pis, Im a developer and work with Linux, python and i am learning now about ROS. The thing is that open source projects grow faster when you have many people working testing and creating. Having options for developers budget will be a great way to get the community involved. I already started the project with the mindset of integrating my own motors and making the design changes required to make it work. The bad thing about that is sometimes that kind of implementation grows outside of the main project and eventially deviates from the original idea and at the end of the day i will endup working on a separate robot with Reachy looks instead of me pushing changes to Reachy software to make it better. It will be amazong if we could get some recomendations of less expensive options and then we can design the required 3d to make it work. Thanks!

1 Like

My intention is totally not to throw flame at Reachy.

I understand that your objection is more a matter of business philosophy and ideology,
and on that front I couldn’t persuade you. If the members of the team are satisfied with the current trajectory of Reachy, then it’s totally unreasonable to expect a big fundamental rework like I asked. (And I know changing the motors alone would require a lot of legwork.)

But consider the target userbases:
(I think we will be in agreement about most of this).

  1. Prosumor: Unlike with smartphones, cars, etc, there is no deluxe consumer market for robots right now. The moment that market opens it will be absolutely flooded with cheap and high quality chinese products.

  2. Enterprise: Reachy isn’t enterprise. (and by the looks of it, its not intended to be so that’s fine).

  3. Researcher: If I was a researcher and I just wanted to play with grasping I would drop 5k on any kuka/scara arm and some generic end effector I could get my hands on. It would be half the hassle with no assembly required, submillimeter precision and high repeatability.

  4. Teleoperation?: The presentation of Reachy is an open source DIY electronics project, but the website features Reachy as a teleoperation app. I think that may be unintended signaling, or its unreasonable expectations. If teleoperation as a product is a goal, enterprise guys are going to dismiss this just on the looks alone in a heartbeat. (Yes I know that’s stupid, but I watched one of these guys do it the other day and go source a scara arm.)

This boils down to my point. Who exactly is Reachy for? Universities, and makerspaces?
Those are your only current options. It is very niche. If that is the intention, carry on.

However:
I want a robot like Reachy in every household in the world. I assume you would love that too. It is my stance that, with proper thought and planning, the ‘hardware’ cost of Reachy can be cut to at least half of its current price of $8500 US dollars (not including tax).

-as few as possible / No custom PCBs. (avoid this R&D and time)
-dumb cmos cams sourcable for practically nothing.
-No planetary/strain wave gears. Those costs aren’t going down. (hence my recommendation of brushless control, and o-drives, with a beefier power supply)
-less time spent on custom software for throwaway hardware

a goal of maximum output for minimum input.

Anyways, I hope I haven’t offended you. This is a common problem with this type of project. https://openbci.com/ had this same exact problem with the Ultracortex “Mark IV” EEG. Sentdex featured it. Millions of people saw it and wanted one. Despite that incredibly publicity I bet less than 200 people bought one. He and and openbci responded with offense when the vast majority of people were saying the price is too high. They said “You don’t respect engineers time.” You know what? They are absolutely correct.
People don’t respect engineers time. Consumers don’t have to respect you. You don’t need their respect. You want their money.

OpenBCI is surely regretting that move at this point, and I noticed the cost has come down to around half of what it was a year ago, so it is at least a goal of theirs.

Please make Reachy cheap enough for me to buy one. My wallet is out.

love your vission, I think that make it cheap and lets make it entertaining and sell upgrades for the people who want the pro version. I think if you make a base model cheaper or printable and parts are accessible then the community can start to add modules. I see a project like this with a lot the community can bring in hardware and software.
The community can add a vaccum cleaner module, a drone extension, maybe make a flying version, legs or wheels, you name it. It looks like a great proyect hopefully Richard Stallmans picture has a real meaning.