18. Open Source Robot Dog, Kickstarter, and Home Robots, with Afreez Gan

2022-05-03 · 1:27:24

In this episode, Audrow Nash speaks to Afreez Gan, who is the founder and CEO of MangDang; MangDang is a Chinese startup that makes Minipupper, an open source robot dog that uses the Robot Operating System (ROS). Minipupper was inspired by and built with lessons learned from the Stanford Pupper, an open source robot dog. In this interview, Afreez speaks about Minipupper's hardware, software, and documentation, about manufacturing Minipupper, and about his eventual goal of building a home robot dog.




  • 0:00:00 - Start
  • 0:01:07 - Introduction
  • 0:04:21 - Origin of Mini Pupper
  • 0:11:27 - Mini Pupper’s hardware
  • 0:22:59 - Minipupper’s software
  • 0:29:08 - Minipupper for education
  • 0:35:19 - Making high quality documentation
  • 0:39:03 - Using ROS 1
  • 0:41:57 - Assembling Minipupper
  • 0:44:17 - The challenges of scaling up manufacturing
  • 0:54:24 - Kickstarter campaign
  • 1:05:22 - Future of Minipupper
  • 1:10:41 - Consumer robot dog
  • 1:16:17 - Manufacturing in China
  • 1:20:18 - Funding
  • 1:24:38 - Future of robotics
  • 1:26:23 - Links


The transcript is for informational purposes and is not guaranteed to be correct.

(0:00:02) Audrow Nash

This is a conversation with our friends who is the founder and CEO of MangDang. MangDang makes an open source robot dog aimed Minipupper, that's roughly the size of a shoebox, and runs the robot operating system. What I think is so cool about Minipupper is that it seems well positioned to lower the barrier of entry for a good robotics education. I think this is true for two reasons. Many poppers inexpensive, yet quite capable, so you can do and learn interesting robotics things with it. And many popper has a community around it, both with the robot operating system for software, and as an open source hardware platform with a good number of users. This community is important so that people can turn online for help rather than requiring an in house expert. I'm Audrow Nash. This is the Sense Think Act Podcast. Thank you to our founding sponsor Open Robotics. If you enjoy this episode, please consider subscribing. And now here's my conversation with Afreez. Hi, Afreez, would you introduce yourself?

(0:01:09) Afreez Gan

Hi, Audrow. Sure. I'm Afreez. I'm the founder and Product Manager of MangDang company.

(0:01:18) Audrow Nash

Yeah. And what do you guys make at MangDang?

(0:01:24) Afreez Gan

Yeah, MangDang. When you set up a company, you have to give a name. Finally, I left my hometown. MangDang is my home town.

(0:01:39) Audrow Nash

Gotcha. That's a nice name. And you make the Minipupper? Would you tell me a bit about the Minipupper?

(0:01:48) Afreez Gan

Yeah, Minipupper is is a ROS, OpenCV, robot dog platform meant for education. actually come from Stanford, the Pupper project. Another very awesome. Open Source project. Yeah.

(0:02:13) Audrow Nash

And what does the Minipupper look like? I think you have one there, but just for our listeners.

(0:02:19) Afreez Gan

Yeah. This is the Minipupper. It have LCD display.

(0:02:26) Audrow Nash

Yep. So it's, uh, oh, go ahead. Yeah, I can describe it if you like, from how I think it looks. It's it's kind of like a little Spot from Boston Dynamics. Like a little quadruped, robot dog. It has. Oh, yeah, go ahead.

(0:02:50) Afreez Gan

Yeah, you know, you maybe I think most of the guys are already see the demo from Boston Dynamics spotted me. And as you know, most of the guys that want to own their own robot that was to explore. Definitely. But you know, you notice that the pricing is so high and it is hard to to accept it. So. So we wanted to make a similar meaning robot everywhere. All right. Exactly.

(0:03:28) Audrow Nash

Yes. And then one thing that's very different between the Boston Dynamics one, and Minipupper is Minipupper is very small. It's like, how long is the body? Maybe it's 30 centimeters or a foot long. And it stands out maybe? Ah, okay, so eight inches long and it probably stands What 10 centimeters high, or 15 centimeters high. Okay. 1515 six inches high. So it's really cute and then has the LCD or the display on its front. So you can have eyes and this kind of thing. Oh, yeah.

(0:04:12) Afreez Gan

We can also add later and camera sensor to those.

(0:04:21) Audrow Nash

Yes. And then so how did Minipupper come out of the Standford Pupper? How What's the origin of Minipupper?

(0:04:32) Afreez Gan

Okay, is a story beginning 20th 20th At first I was only a fan Oh was Stanford a pupper project and sometimes I will answer some a normal question. He is a group for other guys. Other new guys. And then we made made a good friend of his amazing car. As the project owner, and they has Stanford a paper vendor, we ship to many staff for the keys worldwide, and have just started global to head with sample paper, we've got a ton of feedback from as the customers, what do our customers want in their products? And What Kind Of Products Do They like? We learn that a lot of our crew lessons and how to make a robot after more than one years, we will release the manufacturer product. Yeah, actually, and help that brought us to the side by staff for the PEPFAR and the Enlightened collaboration with anything out so really no creator of a snapper Pupper.

(0:05:54) Audrow Nash

How is it different? How so Minipupper is a little smaller than the Standford Pupper? Right, Mike? Is there any other differences? So the maybe the Standford Pupper is I don't know it stands a foot high or something like this or 30 centimeter side

(0:06:10) Afreez Gan

is it bigger than that? Yeah, about all this searching certified I remember 35 or

(0:06:20) Audrow Nash

35 Okay, 14 inches I something like this, okay. So, how else is the Minipupper different than the Stanford popper you made it so that it could be better manufactured or? Yeah.

(0:06:37) Afreez Gan

Actually, as I said before, we delivered many several hangers Stanford a Pupper device and it got a market cap and we know actually, we find Sen staff at a Pupper project he is awesome project manager they are also some problem. So we want to fix them for what kind of problems Yeah. For example, about assemble or somebody we actually optimize something or some point for example, way more we're waiting here. So please yeah, to make it a more easier to assemble and remove eliminate something wrong. This is the assemble and about as a calibration staffer Pupper are really no project we should use a command line tool calibration. Yeah, in Nicaragua, we emplacement CUI tool and the user can use it very convenient. And just not the same point is we make a smaller much smaller to make it a more safe security, you know, when you're debugging a robot, sometimes enemy the out of control exists of performance is so powerful, it may be a security risk. So, we make a huge especially for the children Yeah, for sure we make it much smaller. And it is said even for children and adolescents software to your point that is a sample pack for our project is based around leanings we it is good for for programmer at the end of February we Haussler the software you know we supply is open source for the last SLAM navigation and the Open City functions people can use data to do a much more use case and they the document way especially opposed actually when when they think How are released Standford Pupper project I began I began to follow this piece of project and by all those components the firm's USA and studied as a project at first. The source code is now stables document that is not very well but now it is okay. But actually I remember I used it on last two months tools to rise, staffer the Minipupper project. So I think if we're, if we want to release a project, we hope our user can use it very easily and very quickly. For example, when they do they have several hours, they can assemble it and enjoy it. So that way we supply more document or user guide. Yeah.

(0:10:34) Audrow Nash

Gotcha. That's nice. I mean, if it has poor documentation, and the source code is tough to go through, it can be such a bottle neck to like actually getting anything done with it or trying things on it. And then the fun robotics project can be very difficult. So this sounds really nice if you improve the documentation and give it ROS interfaces and things like this.

(0:11:00) Afreez Gan

And saying they were also held as pre immediate, pre pre compile that I need. People are fo user, customer don't want to compile the source code for several hours so they can download the image. And the flesh is that as the cat, okay.

(0:11:23) Audrow Nash

So you can set it up more easily. To have it all. Wonderful. Yes. Let's see. So I'd like to get into I would love to talk about the software. But I'd like to get a little bit into the hardware of the mini popper. And like what it has on board. So how you first off, it has a Raspberry Pi for the computer, right? Yeah. Yeah, the main controller is the Raspberry Pi. Okay. Yes. And then

(0:11:58) Afreez Gan

that's also based on rest path.

(0:12:04) Audrow Nash

Yes. Awesome. And then then you probably what do you have like a motor driver board or something that the Raspberry Pi controls? Or how does it How does it work? And then it's connected to servos?

(0:12:17) Afreez Gan

Yeah, we actually we cut some much. Cut them about connected to earthbath.

(0:12:27) Audrow Nash

Oh, good Arduino. Is that Arduino? I can't quite tell with the

(0:12:33) Afreez Gan

Yeah, we connect all the servers to our cutter MOBA.

(0:12:38) Audrow Nash

Ah, or your custom board? Oh, okay.

(0:12:43) Afreez Gan

Yeah, this could be the because this is

(0:12:52) Audrow Nash

Yep, there's a Raspberry Pi. Custom board. I mean, yeah, just pop right in. And that custom board has a circle missing from it. And I guess that's for some peripheral or something that the Raspberry Pi that has that sticks through it. That's very cool. Yeah. So then, you how many servos? Does the minibuffer have? 12. Gotcha. And so for each leg. What's so I assume that it's three per leg?

(0:13:30) Afreez Gan

Three steps.

(0:13:32) Audrow Nash

You're so prepared with all the different pieces for explaining it. Love it.

(0:13:36) Afreez Gan

Yeah. Okay.

(0:13:39) Audrow Nash

So you have one. So you have the, basically that elbow of the leg, the shoulder of the leg, which moves one direction and the shoulder also moves. Let's see. So you have two degrees of freedom in the shoulder. Yeah, you could have done it. Very cool. And these so they're servo motors. And you custom built them right for this for the mini bumper.

(0:14:06) Afreez Gan

Yeah, yes. Actually, we don't have to cut them a hater, but we have to, because at first the we buy all kinds of ministers from the market, but we don't, we don't find that anyone can meet our requirement. You know, if we, we, we we want to design a mini size Dhokla dog and we'd have to use mini mini serve as small as possible. But we pies on the shelf. Sir. We cannot find anyone can meet our requirements. You know, we can sometimes times we assemble it and arrive our robot several minutes or several hours. So we'll book break. Yeah. So we, we, I last week where we spend, I remember above the three months to work with our factory with our sub factory to cotton was a syrup. Yeah.

(0:15:32) Audrow Nash

Gotcha. And you made them? What was the major challenge in making your own one? Like, what? Why? Why couldn't you use the existing ones? They weren't small enough? They weren't reliable enough. You said they break frequently? Is it just reliability?

(0:15:49) Afreez Gan

Yeah, yes, it will. If you want to make cameras if you have enough party to add them. Because if you want to make a custom component, the factory will ask your forecast and your budget. Yeah, if you're in a party, that there is no problem. But actually, we have not much value that at first. So we have to reuse as many as their components and also meet our requirement.

(0:16:29) Audrow Nash

So you're saying the factory wanted to reuse its existing tooling to make the servos? Or do I understand, right? So you can keep the price down? Because it's already what the factory has.

(0:16:44) Afreez Gan

Yeah, but we, we could work with cell factory and replaced the motor inside, we use the best motor and we can can we adjust some parameters? Control power inside is us. So Mita

(0:17:09) Audrow Nash

Gotcha. Yeah. Yeah. And the motors, their servo motors, so there's no series elastic components? You know what I mean, with that there's no spring in series with the motor or anything? Or is there?

(0:17:23) Afreez Gan

No. Agenda thought? Or the inside is a third was a very small motor. Yeah, yep.

(0:17:32) Audrow Nash

Is there? Is there a spring in the motor? Or do you because you know how with some of the legged robots, they'll do series, elastic actuation they have a spring in parallel or in series with the motor. So it can deform and deal with the heart impact of the motor hitting the ground or of the feet hitting the ground

(0:17:55) Afreez Gan

that way? Yeah. Are you doing wait? Yeah, when we run the robot, there will be there will be very hot inside. So, so we have to, we have to find a solution. How to Win way. So how to how to make it work stable. And what

(0:18:20) Audrow Nash

do you do? Are you just downright solution to that?

(0:18:27) Afreez Gan

Yeah. Actually, as a sort of factory and as to learn how to test and also just some parameters about the control board inside sir.

(0:18:47) Audrow Nash

Okay, what kind of parameters? What did they adjust? So is it is it basically how far the servo will try to maintain an accurate position or what, like how much effort it will put into that I'm not terribly knowledgeable about servos and controlling them.

(0:19:06) Afreez Gan

Yeah. Actually, I don't remember very clear, I think because they are about to two centimeters, and one is way just as parameters and another solution is the way they think it they replace that much better motor inside and other mechanical optimization.

(0:19:36) Audrow Nash

Gotcha. Now, to make it so you can handle the heat and make it suitable for a legged robot. Gotcha. So is, is Minipupper is it basically, you have the control board in the middle and then you have a way of communicating to the servos and driving power through it. But otherwise the legs the legs are just the server was connected. Basically in order, there's no springs, there's no soft components, there's this kind of thing. And it works because it's a small little robot. Right? Or is that? Is that a fair way to describe it?

(0:20:15) Afreez Gan

Yes. You mean, whether it is there any feedback about a position feedback about the chart feedback? I'm

(0:20:22) Audrow Nash

sure there's position feedback. I mean,

(0:20:27) Afreez Gan

what is this for? Now? What? Now? Is silver fair? There are no any feedback. Wow, it's all

(0:20:40) Audrow Nash

open loop.

(0:20:42) Afreez Gan

Yeah. But you will know if we use we always balances the price and as a performance Yes. How also used today by just with many with many feedback, but a surprising Yes. So hi. Oh,

(0:21:09) Audrow Nash

so your servos don't even have feedback really? Yes. Now, what I thought I thought servos were all closed loop. But I guess if you

(0:21:22) Afreez Gan

if you if so, now, I mean, you have our country control algorithm is the open loop control, it is simple. But it is also if you want to do some deep play algorism it is it is hard to because there is the open loop is on extraversion we are designing our next quarter where we will analyze the feedback but as I said before we we have to balance the price and performance

(0:22:08) Audrow Nash

what is the mini popper now Yeah,

(0:22:12) Afreez Gan

about $500 for the kid gotcha yeah don't put any also is in our next reversion we will add his position and feedback with not much high price almost the same price. We find a simple way to get the position and feedback.

(0:22:40) Audrow Nash

Very cool. Okay, when do you think that next version will come out? As you just had this current version, I think

(0:22:49) Afreez Gan

hey, this is kind of our our next version and we hope to we can release eight into outline

(0:22:57) Audrow Nash

wow so fast. Okay, yeah, that's awesome. How do you how do you get the robot to walk do you use is there like an open source robot legged locomotion lab or is our open source open source library that you can use or did you write it from scratch or how do you do your control your legged locomotion control?

(0:23:25) Afreez Gan

Yeah, this is another story at first you were at a frosted away our software base roster for the Pupper software we have we have to work software, why is the base work it is a similar stamp at the Pupper paper okay and the other version is loss worship. Again the nose warfare we have to implement the framework to control the lag very common framework for the wheel robot, you know there there is a turtle bot platform many guys Yeah, it is a very common framework for the wheel the project that badger April, we if we want to implement as a lag robot based on Joe's platform, we have to design and implement has the framework at first we I asked several technical guys and it is has almost I remember almost two months, how to design it or how to end Dementor How long do can we implement as a laughter? We are very lucky, we found there are already a good project, you could have another open source project. Then Kemper lead by Biogen, who is located in Singapore. We are now we are very good friends now. And during our project we he also give us a much have. Yeah.

(0:25:39) Audrow Nash

Wonderful. So, you found an open source solution you were able to use with ROS and that gets you most of the way did you have to extend it or did you put code back into the original to improve it? Or do you just use it as is

(0:25:56) Afreez Gan

actually way way it is not only just the integrated way which we are based his project that by the way are also double confirm with you and discuss with you and how to use data. Much better. And you Pupper? Yeah.

(0:26:17) Audrow Nash

Yeah, you optimize it from there. So it works really well. Yeah. Awesome. Okay, how did you? How did you optimize it for Minipupper? Like I've seen some things where they would have they for you know, the robot soccer tournaments and they'd have a little now robot go and try to kick the ball. I would they were trying to make them the now robots really fast. I think it was now actually, maybe I may be wrong. I think it's not an hour robot. I think it's some other humanoid robot. But what they did is they would set them up and have them race back and forth with different parameter sets. To try to find the fastest parameters. How did you? How did you get it prepared for mini popper? Did you do something similar? Where you're racing back and forth? Or how did you find what works really? Well?

(0:27:13) Afreez Gan

Yeah, our target mini perpetuity is for the education, but not for the race competition. So we have to make it as simple as possible to make other guys to study or catamaran if we have a freezer robot, because it is a very small robot dog and component, the surf is is a low cost result is the performance is okay for me Pupper patter. Eight is a view want to write very quickly. Maybe it will not as stable. Maybe. Yeah. All over. Yeah. Yeah. But if it didn't write up, I don't have enough. You cannot use the leader ADCs encoder for later to do your system and navigation because it is another stable and Amanda,

(0:28:18) Audrow Nash

it's shifting shifting. Yeah. keeps breaking its map

(0:28:21) Afreez Gan

to build as a map. Yeah. Yes, because

(0:28:26) Audrow Nash

you use the planar LIDAR. Right? On many bumper. So it's the LIDAR that shoots straight out. And so if you have the plane going out, and it's tipping, tipping tipping, and you're trying to use that as your map, that would be really difficult, I imagine to build a world map.

(0:28:44) Afreez Gan

Yeah, well, you could imagine and make a that around more stable, stable, they're able to do

(0:28:53) Audrow Nash

so stability was kind of the big thing you optimized for when trying to make Minipupper walk, not like speed or anything because you're interested in it for education. And that means like, can you learn with this as a platform? Yeah. Yeah. And so then you can try learning about Lidar and whatever other sensors, your cameras. Okay, yeah. Tell me a bit more. So, is this robot already being used in education and where and how is it being used in the curriculum?

(0:29:29) Afreez Gan

Humans education, right. Yes. Yeah, yeah. deoria we deliver as Stanford a Pupper our units, we, we found many. Some of the user come from the university, most of the university. Yeah. Even some PhD students. Yeah. And they, yeah, they hope they they can they looking forward acceptable a robot have a platform to test their algorithm, maybe as video algorithm or other AI algorithm. So wait, you have the with. And now, as I know, some, some universities are trying to, to use me Pupper for their lessons. And we are also discussing with how to improve MiniPro. And what they are looking for. Yeah, I think it is a long, long journey to improve the education platform. Yeah.

(0:30:57) Audrow Nash

So it's mostly targeting universities, not like high schools or middle schools or anything like this.

(0:31:07) Afreez Gan

Yeah. Actually, at current our, our level now, the most students come from the university, but they are also some come from the middle school, high school. Okay, if we,

(0:31:26) Audrow Nash

yes, just starting? Yeah.

(0:31:31) Afreez Gan

Because it's not easy.

(0:31:35) Audrow Nash

Yeah. Yes, for sure. Especially I imagine going from university like starting with grad students, and then going towards maybe undergrads and you teach them during your education. And then like, the further the lower and age you go, are there like the lower and education level? I imagine the harder it is because you need to develop a more structured curriculum, I would imagine. Yeah. Like, with the PhD students, you say, here's some documentation on how it works. And they go off and do things. And then with the with the undergrads are doing their coursework? I imagine you still have to put together a little bit of a curriculum or work with professors or technicians at the university to put together a curriculum. Have you done much of that, where you're putting together a curriculum around Minipupper and the kinds of things that can do?

(0:32:32) Afreez Gan

Yeah. Now, we are, we are just supplying the proper answers there. based platform. You know, your university students have the technical background. And we only can supply the the general user guide and the time period. And if they have some problem, we can they can say No, for me, I

(0:33:05) Audrow Nash

come to you.

(0:33:06) Afreez Gan

Yeah. So we kind of suppose that. Yeah.

(0:33:12) Audrow Nash

Gotcha. Any plans in the future to possibly help with the curriculum design around the Minipupper? Or is it kind of your documentation is in a sense, the well it is the documentation you're planning on using or not? Not necessarily attaching a course to it?

(0:33:32) Afreez Gan

Yeah, actually, we were planning how to make a detailed document or user guide, where we hope to make a a real education platform, not only for the university is to say that we can also hope we'll also have the middle school high school students to learn understanding. We can make a detailed document for all level users.

(0:34:13) Audrow Nash

Gotcha. Have you started that? Or is it it's a future efforts?

(0:34:18) Afreez Gan

Yeah, we are. Where are you in? Yeah, we have online online document that we will add as a additional part, standby snap. Yeah, we hope we can. Finish is a document we see as three or five months.

(0:34:46) Audrow Nash

Ah, very exciting. And then maybe you'll see more universities using it and it will grow from there. That'd be very cool to have a very nice, inexpensive legged robot like how cool I mean In my robotics education I was using lucky to have the thing, but like little tiny robots that had a lidar on top and didn't navigation, and legged robots are just so much. There I am. I'm so much more interested in legged robots than wheeled robots. And so it's really cool if that can be in the education program. Yeah. How do you how do you work to make really good documentation? Because I think a lot of like, many of the robots I've used that have come from factories, have had pretty terrible documentation. How do you try to make good documentation any, any ideas on what you can do to make it as good as possible? Or?

(0:35:43) Afreez Gan

Yeah. And of course, we have many improvements of the document. And of course, the we made our first version document, and user will give you many, many feedback. And the user can maybe make their own User Guide. For example, when it is the many hundreds of a device to Japan marking and sun. Japan makers that make their own token are very detailed a document, so I

(0:36:23) Audrow Nash

will put it back in.

(0:36:26) Afreez Gan

Yeah. So we actually a lot of money is not only come from us, but also come from all the guys, all the customers, we are cool, and contributing to the Terminator data, and making it more and more better.

(0:36:48) Audrow Nash

Gotcha. Very cool. And so you mentioned a little bit, you can put a lidar on top of the menu upper. I think you can also put a what is it? Is it an oak D? 3d camera?

(0:37:02) Afreez Gan

Yeah, open CV? 3d camera module? Yeah, yeah. So yeah, ideas. Lead by branding, branding team. Branding is open CVA, a key architect, architect. And, yeah, he's also the CEO of Orlok. Sonics company, located the AAMC. Okay.

(0:37:29) Audrow Nash

Gotcha. So, you put these sensors on, and these are so that people who are interested in playing around with it. So whether it be students or I can even see hobbyists being very interested in playing with mining like I, for example, I'm interested in playing with one of the many buffers, but so you can put the sensors on it. And with the camera and the LiDAR, you can build a map. And you can do a lot of different robotics things. Because you can have the robot map its space and move around in the space and do some understanding of the scene or the environment. That's kind of the intention with those to enable it so people can use them, or, yeah, it's quite cool.

(0:38:11) Afreez Gan

Yeah, you're right, we've put later this place and we can put his camera at this place. Yeah, we

(0:38:23) Audrow Nash

do that later kind of sits on the back. And the camera sits a bit further forward, kind of on top of the body.

(0:38:30) Afreez Gan

Yeah, we also, we also supplies a user guide for the leader. And as a 3d camera module, you also can look at our guide, and now is only the basic guide. And user can use our guided tool, Ryan as the sensor. But if a user wants to implement that some use case, they have to study and implemented their use case

(0:39:00) Audrow Nash

themselves. Gotcha. And so this is and it's running the robot operating system or ROS is it running ROS 1 or ROS 2?

(0:39:10) Afreez Gan

I just one

(0:39:12) Audrow Nash

plus one. How did so see? How did you decide to Well, for us one was the Stanford upper running was the Stanford PACCAR running ROS 1 are you brought over ROS 1 for this and implemented or found that one package that worked really well? For the Lego

(0:39:34) Afreez Gan

auction? Yeah, we've built a real city environment. That's our Pupper. Only is the product leanings. Meaning so ask.

(0:39:45) Audrow Nash

Okay, so this was you were saying you found that one repository online that did a lot of the legged locomotion for you the control of the legs for walking? And that was ROS 1 And then you've kind of Gotcha. Do you have any plans on moving to ROS to? Or is is ROS 1 the good one for now?

(0:40:09) Afreez Gan

It was Ryan's work very well now. But till we we are also discussing on several time how to update the tour ROS to version. And we hope I think I didn't need a lot of work to update updated. Yeah, and I think we have still not very clear plan to update two rows for but we also, we also hope to update. Maybe I hope we can update the two rows to people the end of this year. Yeah, we can update this version in about and I want that robot.

(0:40:58) Audrow Nash

Yeah, it seems like a lot of companies are kind of feeling this pain of transitioning from ROS, one to ROS to

(0:41:05) Afreez Gan


(0:41:08) Audrow Nash

There is something if you use the Python client library for ROS. One, there's a thing that it's underscore ROS py. And you can import it and it just uses ROS to but it keeps all the same ROS 1 interfaces. I thought that was quite cool. Okay, I don't know if that would ease the transition. But there I don't think there's a c plus, like, I don't think there's a C Plus Plus client library analog to that at the moment. But it's a cool idea to just switch it out. So you can use it. And then you can gradually replace things. As you go from their

(0:41:48) Afreez Gan

way I try our best to move forward step by step. But yeah.

(0:41:55) Audrow Nash

Gotcha. Let's see. So when a user orders one of these mini poppers, it's a kit and they have to assemble it themselves. Is that correct?

(0:42:09) Afreez Gan

Yeah, we we supply several, several stos. One is the basic Acade. You know, you know, some, some some makers, they already have their resume bought, they don't want buy additional box. So they only want some some components. So with supplies, the basic Okay, and not include them. And then some, I said I, I hope to buy all the components, and I can assemble it easily. Yeah, sometimes. I hope by is a pre assembled unit. And after I get it, I can see jointer right now. So actually, yeah, we supply three eSkills.

(0:43:10) Audrow Nash

Gotcha. That's cool. And the body of it is 3d printed. Right. And the legs as well or several of the components are 3d printed. Right?

(0:43:22) Afreez Gan

Yes, this is one of the challenges for our, to our product. Yeah. We are 3d printed. You know, if you if you only make a prototype, 3d printed, there is no problem. You can print it with several hours or many hours. Yeah. If you want to mass production, for example, we I traveled to shinjang. Our lender? I asked them, I need several 1000 units. 3d printed, it is hard to find is the vendor. Yeah. Wow.

(0:44:20) Audrow Nash

So let's see. And the trade off is that you can 3d print and it's really good for prototypes when you're doing like one here and one there and learning. But mass production if you were gonna make 100,000 of them, then you go and you get the molds done. So you can like cast the parts or something and it's very efficient, but it's very expensive upfront, right. But so that and many popper is kind of in the middle where your volume is not incredibly large. It's not like hundreds of 1000s It's like 1000s Perhaps. And so that's why you're thinking 3d print?

(0:45:02) Afreez Gan

Yes. You know, for for some part? Well, yeah, you'll see for some yellow pattern, they actually, at first that house components are 3d printed. And then it is harder for us to find the vendor who can supply so many 3d printed 3d printed parts. So some of the some of the path we actually now is load from both. And that sounds really balanced. Let's

(0:45:41) Audrow Nash

see. Yeah, the cost. And the end of the day. Gotcha. And the so maybe for the next model that you think will come out, you said like June or July or something, I think that one, are you moving more and more towards molds for everything? Like, grab, try to reuse as many of the pieces as you are already having molds for? And then eventually have it be more? Yeah. So

(0:46:09) Afreez Gan

we, yeah, we'll reuse some sums and both paths? Yeah. Yes, we have. Maybe our natural working is, you'll see, we can we can remove this path, rest path. This path? Yeah, we can have a very good platform, you can. So

(0:46:36) Audrow Nash

if you remove kind of a bump in the front, then you can have a big flat space to put sensors and whatever it is you want. Yeah. Awesome. Let's see, would you so we're touching on it a little bit. Would you tell me a bit about the manufacturing process of this? And so you were saying before we started recording, you go down to Shenzhen, and they work on like, it's easy to prototype? And just tell me a little bit about the manufacturing down there?

(0:47:07) Afreez Gan

That's yeah, you know, yeah. First, we, it is not easy. For the hardware. Hardware device, especially and now was now situation, you'll have to you'll have to powerful supply, supply chain capabilities. Otherwise, you know, you're you can talk as your opponent is a good, but the coolest thing is, actually I have several years of hardware products. And I also have some resource supply chain resource, so we can get get our components. Yeah. Yeah, we also, you know, we use the rest, Pat. And there is also very hard to get the book.

(0:48:14) Audrow Nash

I've heard Yeah. Yeah.

(0:48:17) Afreez Gan

But we are so lucky. And we were I'm wondering and thinking about how, after our campaign, and we have to prepare the production, the most important commodity is Rat Pack. Before we can buy merch by either name, and price, after end of our campaign, it is hard to get get as much fat. But we are so lucky. And we bet Westpac to like our product. Wonderful. Yeah, we are so lucky. And so I connected with Sylvester talking about we needed less hype for our product. And so, actually, we've got enough respite from from ROS back come there. UK. Yeah. Okay, so

(0:49:25) Audrow Nash

it's so funny. Yeah, the part shortages have been really tough, especially for hardware companies. And especially like, I don't know, I hear stories about companies ripping out Raspberry Pi's from other devices so that they can use them. So like buy a $200 thing and then rip out the Raspberry Pi, which was like $40 Before this whole supply shortage. Gotcha. So it's really nice that you have the ability to keep getting Raspberry Pi's. Let's see. And so

(0:49:59) Afreez Gan

here's a theory As a manufacturer, the leading time is another risk because of

(0:50:06) Audrow Nash

the waiting time, so parts what is the leading time

(0:50:12) Afreez Gan

while the story is at first we have to we have a tool release also our device how that came to our backers before the end of last year during during I poacher our effect, I also look at our factory last year and push that progress. But you know, last year shinjang as COVID COVID-19 situation is not very good. And we already have we have to delay a shipment. So it is it is very hard to control. Accurate deadline for the shipping for

(0:51:02) Audrow Nash

the whole world. Yeah, especially. I mean, it touches so many places in so many industries, the raspberry pies, I don't know where they're made, but you said it's a UK company. And then you have the Schengen area. And then there's the whole global pandemic situation, there's parts and materials that need to come together. So yeah, I could imagine that it would be very difficult to predict the lead tie for all the parts.

(0:51:28) Afreez Gan

Yeah, yeah. fracture, I really hope many thanks for many guys for Okay, well give our madhhab ROS, a Mac Suissa. Team give us and the hardware.

(0:51:50) Audrow Nash

What's the what's the process when you go down to Shenzhen? And you have an idea that you want to get made? How does it work? You go to different companies and or different factories and you say I want to make this and then they work out what they can do? Or how does it work? How does it work to manufacture something?

(0:52:10) Afreez Gan

Yeah, actually, it is not a lot of hard work and tissue work. Because if you know you want to, for example, I was cutting myself, I don't know first. Just terrible kind of work. But it is a to cannot work best. So I just leave. I try to leave a hotel very near that as a satisfactory. I really, I went to their factory, and it works out and then push them.

(0:52:46) Audrow Nash

Oh, wow. That's crazy. Yeah. Yeah, I

(0:52:51) Afreez Gan

said, we have to find a better solution for the better performance. At first it is possible, but a real person day and day. Finally, we found a solution.

(0:53:08) Audrow Nash

I've heard that a number of times where like, the CEO or CFO or whoever is responsible for manufacturing will go and just basically stay with the factory and keep pushing them to try to make refinements and improve the quality or get it to work in some way. This kind of thing.

(0:53:30) Afreez Gan

Yeah. You know, we you know, we are a very small startup company and before the campaign layers. I was the only guy the full time because other guys I promise. And

(0:53:52) Audrow Nash

how large are you now?

(0:53:54) Afreez Gan

Yeah. Now we have several full time guys. Yeah.

(0:53:59) Audrow Nash

Gotcha. Several, like five or several.

(0:54:03) Afreez Gan

Right about Friday for full time and several hours that are part time.

(0:54:10) Audrow Nash

Guys, gotcha.

(0:54:11) Afreez Gan

We'll tell you about 10. Guys.

(0:54:14) Audrow Nash

Yeah, about 10 guys with full and part time I see. Five Guys. Gotcha. Okay. Yeah, that's it's, uh, you were the only full time when? Before the campaign and so the campaign. That's a Kickstarter campaign you're referring to, right?

(0:54:36) Afreez Gan

Yeah. Yes. That campaign is not have a funny, interesting story. And I would love to hear that. But let me share your story. Here's our last story. All right. Yeah, at first, you know, back to 2010. Here we are. The term for the Pupper vendor, and a co worker is anything how. And after we release the Minipupper, a Nathan Kao also told me Pupper onto their staff at the Pupper website. And I think maybe we can seal or deliver Minipupper just like snapper Pupper. They put that information on their website. It was a category one, get a good visibility. Yeah, they can send a mail and pay by by paper or other online payment. Yeah, so I have no, I have no any crowdfunding campaign knowledge before. Before Ming Pupper, but some day, nice. And Tao said, why not try campaign? I said I have no any background. Maybe it's impossible or difficult for me. So some days later, they say send me a same method yet again. And again. Why not try campaign? I Sunday. Sunday I have was free time. And two was was a media campaign. I started to study how to make work campaign. Yeah. Yeah. So I'm not very clear. You know, it's hard for me to make it clear. And we I made a decision to do a campaign. I thought whatever the result, issue shouldn't be very interesting. Another another increase in DNS for me. So it is for sure. Yeah, if you decide to campaign, you have to prepare all kinds of lawyer all kinds of lessons and prepare all kinds of work. And I also double confirm with some lenders or some friends, ask them what how to do the campaign. Some questions you can get to the answer easily about the campaign process, by the some standard question is not easy to find his answer. Because if you want to do the campaign, you have to prepare the PR video, the Samza pictures and the campaign pay design and all kinds of work. So at first, we discussed with several vendors, how to do the PR video, the if you have enough data, you can do it and use it. And our professional. Yeah, lots of professional team connect you and then they can do the QA video. But yeah, the budget is, for example. 20k dollars. Yeah, but but not, but at that time at the different code for me to decide. Because it's a big investment decide Yeah,

(0:58:35) Audrow Nash

especially if it's just you, for sure.

(0:58:40) Afreez Gan

And the last we I asked the my friends brain team to take us to record a video. But the unities is before before one week before the campaign. The guy who happened to make his PR video heard his hand and he cannot edit his video.

(0:59:12) Audrow Nash

Oh, no way.

(0:59:14) Afreez Gan

Yeah, we haven't prepared the voice because we have to prepare the video and the voice. I lost that. Yeah, I lost. I finished as laughter snap finals. Voice and integrators combines video and audio. Yeah. Gotcha. Yeah. So a lot although the final PR video is not so professional, but it can work it's quite another Yeah, another another very interesting is you know if you're a If you want to make a successful Kickstarter campaign you will have to God's Kickstarter campaign official platform project we love label. Ah, yeah, yeah. Yeah. But even but, you know, we I have no any other. No any I know how about the campaign? The question for me is how to get the project will have a label? I asked. I asked another guy hired me to do the campaign. At first before, even before was one week before the campaign I asked him when we can't get as a project, we love labor. He said, If your project is good enough, you will get it. Yeah. But this is a Das? Is that the answer I want to get? So I asked him Well, time and time and time again. And it is harder to have data that you like to answer. But I'm not alike. We can just retweeting the labor so I'll try to find the picture campaign contact window and I want to African chrome with the official contact window. So I the last time that I found contact window hazard here she is Heather and I was at design tech category our pigs trap. And here's the icing I sent her a male introduced our programming power project and how to book a video conference. I remember one Saturday afternoon so before the surprise happens our video meeting is before the meet here before the video meet him ice and we already may we already made campaign paid but it is very rough. Very proper. The video is not to the fixed version. Some pictures are not the fixed version looks not good. And is

(1:02:52) Audrow Nash

not very professional. Yeah. The page

(1:02:56) Afreez Gan

actually looks bad. Okay, but But before before the meeting with hedsor Western is a meeting are we saying is appealing to her and introduce our more information our me Pupper our immediate information. Our video is a Sunday afternoon. And we are surprised and added that just the pneumonia we find our project already has already give our valuable project we love the labor. Very surprised. Yeah. Yeah.

(1:03:42) Audrow Nash

That was so nice. Very happy. Yeah, I would imagine.

(1:03:45) Afreez Gan

Yeah, very happy. And the other week as a laborer. I asked the guy who had who support me to do the campaign. Yeah, I we already finished our target that before the campaign was our next target.

(1:04:04) Audrow Nash

Yeah. Yeah. So interesting. That's crazy. Yeah, it was a very successful campaigns. Congratulations. Oh, yeah. Yeah,

(1:04:15) Afreez Gan

thank you. Thank you. Another Yeah. Another interesting is open CV open time away. Branding team will also do their Kickstarter campaign added at a time away away, almost to the campaign at the same time.

(1:04:35) Audrow Nash

Yeah. Ah,

(1:04:38) Afreez Gan

you could see his brand name is Ken Scott is also your

(1:04:44) Audrow Nash

local robotics. Yeah. See, see

(1:04:47) Afreez Gan

traducida brand new to me to link the US together. And wonderful. We are very close to here wonderful. crew together, more and more deeply and granted him are now have already failed at their Docker Hub. Yeah, they have a we'll use our mean Pupper. And we will run it together. Yeah. More in the pipeline.

(1:05:17) Audrow Nash

That'd be great. What do you think? So you'll work together? I guess where? Where do you see many pupper? Going? So you already mentioned that there's another version coming out? And you're putting together the documentation? Where do you kind of see it going? Maybe two to five years from now?

(1:05:37) Afreez Gan

Yeah. In my dream, we hope or we have a tool or product line. Why is education we really hope to release Good, good, local data platform for education. They're everywhere. Not only universities, but also middle school, high school students and other any guys who like as Loboda they can study and enjoy as your product product. And another product line is the home robot. We hope Yeah, we hope or we can release useful and valuable and acceptable home robot, home robot we lag from global dark, consumer, consumer double dark, but you know, it is alien not do that. Because we have to we have to move forward carefully step by step.

(1:06:52) Audrow Nash

What do you mean? What? What are some of the obstacles that you must figure out to move forward? For the consumer robot dog?

(1:07:02) Afreez Gan

Yeah, for the consumer of our dog, we have to make it as a product. As simple as possible. You know, if you have some technical background, we can we can discuss how to calibration. But if for the normal for the consumer product that you you don't you cannot say you cannot teach your customers to calibrate

(1:07:33) Audrow Nash

it. So it has to have like one button like a Roomba and then turn on and do things and also has to be cheap to

(1:07:43) Afreez Gan

hear. Yeah, the price is gonna be very cheap. And as the function to the various stable. Yeah, yes. We hope we are trying our best maybe because the end of this year, maybe we can release our consumer local Target prototype. Maybe later this year, early next year. What do you think? Oh, go ahead. Yeah, we have the study was that year? Yeah. That a the real requirements from the customers, you know, during our staff are probably released about 580. device. And mini paper with we already released about 2k. Wow, this wonderful device. Yeah. And our next version I next. I mean, I named the mini Piper v2, we hope to really say that the two outlier. And I hope we can release a final sound in the outtakes of the device. If you are released. The more device the already is, the more feedback you can get. Yeah, if you're not to the more feedback, you can make it up better and better. Definitely, Someday, you'll find a consumer solution.

(1:09:21) Audrow Nash

Yes, it seems like the Minipupper and Minipupper v2. It's a smart way to one get feedback as you're saying and to grow the number of units that you're making, because the consumer one will probably be fairly similar. And you can refine the control algorithms and you can refine everything else about it. So it's a nice way to go from a small market that lets you keep growing to a much bigger market to grow a lot more, I think sounds like a good plan

(1:09:55) Afreez Gan

to make that fit. Yeah, because such a very cool Actually we Yeah, because we actually we have we have not much a Makita. To to try. We have to areas them very carefully. Yep.

(1:10:14) Audrow Nash

I see, yes, it's a very much taking what you have getting a little more reinvesting it back into your company grow a little bit bigger, and do the same thing, grow, grow, grow, grow, until eventually you can be a big company. Ah, I hope to look back in the future and see you guys be enormous. And very successful company. So with the consumer robot dog that you're imagining, what will it do? So is it is it just gonna walk around? Is there the the human interaction part seems really difficult for that, but what's going to be kind of the thing that people buy it for?

(1:10:58) Afreez Gan

Ya got to dot your point? Yeah. I was asked by many guys the same question. Yeah, yes. In my opinion, in my opinion, it is. I think no one can give a very clear answer. For example, if that answer is very clear. I think,

(1:11:31) Audrow Nash

oh, everyone else has already done.

(1:11:34) Afreez Gan

Yeah, I think a marketer is very clear. And as a product feature is very clear. We have no opportunity for the startup company, the big company will. But for me, I believe you can make it after three or five years later. I think those are the home home robot market that will come. I believe it doesn't matter what kind of a home robot looks like. I believe he should be as a lateral robot, but not the wheel robot. So so so if the if it is the leg robot, we can we can immediately add a four leg or two leg for now, maybe? Finally. I think AI is used for leg because it is more like two animals. Yes. So yeah, I believe was the maintenance. So we, we moved with their target stamp I stamp. But I did that first. Maybe in the end of this year or early next year, our first home robot prototype, or maybe you can think that it is a mobile smart speaker on mobile. Yeah. Yep.

(1:13:17) Audrow Nash

Kind of like Amazon is doing and a lot of these other companies. Gotcha. Yeah, and legs. I mean, like carpet is not a problem if you have to like step up onto the carpet or stairs or anything. Like if you have a legged robot, there's a lot of advantages for getting around the home. Especially if it's a consumer products, like you buy a Roomba. And it can't go in the like little dip in the living room or something like this that has a stair to editor, whatever it might be. So legs clearly have those advantages. Four versus two is an interesting discussion for is a lot easier, I think as a roboticist because you can be passively stable at anytime because you can have three points of contact or this kind of thing. Yeah.

(1:14:08) Afreez Gan

Yeah, you'll think maybe five or 10 years later, a few days. Home robot with four leg you may be feel very good. It is like animals, but maybe you will find as they are two leg like a human in your home. Maybe it is not feel better.

(1:14:29) Audrow Nash

It's like a little human. Yeah, it's a little stressful. I could see that. I wonder I mean, birds birds have two legs, also. So like chickens and things like this. And we're not terribly intimidated by them. But I think a chicken shaped robot would be very funny. Like it's not a design that I see. I think ever but yeah, I see what you mean. And then also easier to manufacture probably point like feet are really hard as far as I understand for you like buy big needle robots? Because like, a lot easier with like sore points, like you're doing with Minipupper?

(1:15:11) Afreez Gan

Yeah, the most of the most time consuming data users use lag. Yeah. Here, yeah, we have a tool. Now what I need, I still need about three hours to assemble it, and additional half an hour to calibration it. And our next version all my Takita is Katherine was, you know, I need more than three hours to assemble it.

(1:15:47) Audrow Nash

And you've done a lot of that's yeah.

(1:15:50) Afreez Gan

But other customers maybe need more time to assemble and enjoined? Yeah, one extra version? I hope the assemble. We can assemble it within one hour. And don't need to any more calibration. Any maybe the auto calibration?

(1:16:14) Audrow Nash

That would be nice. Yeah. Yeah. Let's see. Bigger Picture. I'd love to so segwaying a little bit. I would love to hear your because I mean, I'm in the United States. I would love to hear your impression of kind of the robotics industry in China. And the manufacturing is very interesting. So like, I don't know, how do you how do you think of it?

(1:16:42) Afreez Gan

You me as a robot I manufacture?

(1:16:46) Audrow Nash

Yes. In the Schengen area, and then around China? Just yeah. I don't know how, like for the whole robotics field, how how do we how can we think of Shin Jen? And? I don't know.

(1:17:01) Afreez Gan

Yeah, I think, shouldn't they, the supply chain is a huge one for us in general. And you can find all kinds of vendors. And they can't give you some solution. But if you want to do some innovation, you have to you have to design the Pycelle you can find all kinds of solutions. The common solution, it is easy to find a common solution in ancient in Makeda, and a winner. But if you want to make some innovation product, or you have to design it myself.

(1:17:48) Audrow Nash

Yeah. Gotcha. And so if you do design it, oh, go ahead.

(1:17:54) Afreez Gan

I mean, if you review, product, manufacture a common hardware device, it is very easy. Because it certainly is a very convenient.

(1:18:12) Audrow Nash

Gotcha. So if you say you do come up with something new like you with your servo motors that are a little different than what's been done or maybe I don't know anything else that's new. How does it work? To get a factory in Shenzhen? Making it like do you have to work with them a lot? Do you kind of give them what you need? And they'll figure it out? Or how does it work? Yeah.

(1:18:40) Afreez Gan

Yes, about you can also find many vendors initially market and during a tutorial deliver staff at a Pupper, I already know some sub so vendors. So we asked them one by one and compare their service and as a technology background, and their selectors. The better battle rounds and then to discuss to to give give them our detailed requirement and discuss our solution discusses the solutions maybe, you know, as the forecast. Now the fourth category is to not match and you'll need I didn't not a very easy to to push them to some ketema Yep.

(1:19:46) Audrow Nash

Gotcha. Unless you're making I don't know a million of them or some crazy number. That it's really worth it for them to customize. This kind of thing. Yes, yeah.

(1:19:56) Afreez Gan

Yeah, yeah, I want an extra wash. I will resist that serves, because, you know, if you're wanting to do some innovation or some big change, to make it simple, you will have to redesign some, some key components.

(1:20:16) Audrow Nash

Yep. Gotcha. How, how is the I mean, this kind of two questions. How are you guys funded? And how is the investor climate in China to you?

(1:20:36) Afreez Gan

Okay, so the first question is, how to get the data right?

(1:20:42) Audrow Nash

For you, so are you are you bootstrapping Are you putting your own money in and are you accepting investment from people from? Are you locally funded? Does the government help with your funding? Or I'm just wondering how the funding so far so are you guys

(1:21:03) Afreez Gan

different? i Let's go back to Stanford as a Stanford a paracord project. And at first saw it I used my own money to buy all the components about Africa and and they enjoy it and Alaska, i becomes the vendor of the temple Pupper and the bigger because after we said rotis castle with Nathan Kao by mail or videoconference, Nathan asked me whether it is possible to supply as a kid from China marketer because China supply chain is convenient and as the price is, is lower. So I look around and all the components from the Kenya market and deliver a stamp or the Pupper with with low price. But frankly, speaking of the day, the some profit after deliver several hundreds of staff for the Pupper project we got we got some profit. And I use a profit to developer Minipupper. So yeah, so when the Pupper campaign after the campaign, we got much more funding. And then we use that money to developer our next version. So Ron and Rob Yep. Actually, now or now. Now it is. I I didn't actually i not i did not use the notch in my own money.

(1:23:06) Audrow Nash

In the latest mini popper? Yeah, yeah. Nice. That's great. Let's see. And then, um, do you think going forward? Will you take investor money or try to bootstrap the whole time and keep becoming?

(1:23:24) Afreez Gan

Yes. We have as a plan to get investment maybe later this year. And we are so lucky. And we actually now we already got to the incubator wacky incubator program approval in Hong Kong. Congrats. Yeah, they will. Yeah. They will give us some support and not only offers some funding, but also, you know, in Hong Kong. The VC environment is very, very convenient. Yeah. Yeah, well, we can. Yeah, well, I hope we can get our first round an investor later this year, and then we can use the fun to to look more. More technical guys, and accelerator our, our rd and marketing.

(1:24:34) Audrow Nash

Gotcha. Yeah. Let's see. So beginning to wrap up. What do you think? What do you think robotics will look like? Going forward? So the next like two to five years. Where do you think we'll be in robotics?

(1:24:57) Afreez Gan

Yeah, I'm very positive. Go to the robotics market. And you know, now, vision technology, speak as a voice or re as maker Ria, voice technology. There are many, many solutions. And I think that we are already ready. The only need the only thing we need to do is how to combine all kinds of all kinds of functions or technologies into a total product require the product, and they release it to the customer with acceptable price, and variable use aka. And I think I'm very positive, but I think maybe within five years, three or five years later, I think most people will see more and more robot products. And I also believe, within three years we we can release as home robot top consumer product will be sold in no time, but some company will release it. Okay. Well,

(1:26:24) Audrow Nash

I hope it's you guys. Fingers crossed for it. Yeah, let's see. So we're wrapping up. Is there any links or contact info or anything you'd like to share with our watchers and listeners? Yeah. All right.

(1:26:44) Afreez Gan

Also, talk a minute and a resource of Amitabha is your company data on our GitHub? Link? I think you can get all the all the things on our GitHub.

(1:27:01) Audrow Nash

Awesome. All right. Very good. Thank you. Nice speaking with you. Yeah.

(1:27:07) Afreez Gan

Thank you so much. Yes. Thank you so much. That is amazing. Hi, everyone. Okay. Hi.

(1:27:17) Audrow Nash

Thanks for listening to this conversation with our friends. Thank you again to our founding sponsor open robotics. If you enjoy this interview, please consider subscribing. See you next time.