> BASIC is accessible for free to all users of Innate robots for 300 cumulative hours - and probably more if you ask us.
Is BASIC used just to create the behaviours or to run them too? It sounds like this is an API you host that turns a behaviour like "pick up socks" into ROS2 motor commands for the robot. Are you open sourcing this too, so anyone can run the (presumably GPU heavy) backend?
Does the robot needs an internet connection to work?
Also, more importantly, what does it look like with googly eyes stuck on?
On BASIC: Yes it does require an internet connection and until we figure out how this works for you it will remain free of use!
It is required to run them, not to create them. And it's not about running "pick_up_socks", this one can already run on your robot. BASIC is required to chain it with other tasks such as navigating to another part of your house and then running another skill to drop the sock somewhere for example
Thank you for the remark, we will make it clearer in the docs
As a consequence: The robot does not necessarily require Internet to run, but if you want it to chain tasks while talking and using memory, yes it does.
How complex of tasks can you give it? Seems, not to be punny, super basic.
Can it do complex tasks like "pick up socks from room A, drive to room B, and put in basket"? Is the intention to allow hobbyists to do actual work with it or is this version purely novelty rather than a functional "personal robot"?
Additionally, what is the limitation on speed of movement? It seems very slow in movement, is that intentional for safety or is that purely because of running the AI model locally?
The example you gave is exactly this. The driving around in the example use-cases shows an example of going from room to room and execute policies in context:
I take from your comment that the full capabilities of the robot are not properly represented and I take a note to film longer ones. And it can definitely do what you just asked. And multiple times in a row. I will note that it depends on the training quality of course.
On speed of movement, I now realize we didn't mention it anywhere so I added it in the overview but it's pretty fast. 0.7m.s-1 for the base, and the arm can be tormented quite a bit. Just took this video for you:
Does the MARS hardware really remove the hidden extras (computer with a gpu) mentioned as the downside of HF SO-101 or LeKiwi? While a jetson is good for inference, I feel like to train VLAs you would need access to a powerful GPU regardless. For Lerobot based hardware training ACT was relatively low profile if you use low resolution for the camera feeds, but with increased resolution or with more than one camera I already saw needing more than 8GB of VRAM. If VLA is on the table, finetuning something like the open sourced version of pi0 should already necessitate access to more than one 4090 or above I think.
Also, do you have plans for community-level datasets? I think Lerobot sort of does this with their data recording pipeline and HF integration.
One of our objectives here was to fix everything that we don't like about the SO-101 and the Kiwi, which have several hardware and software flaws in our view. Including, yes, the constant need for a computer to simply run your robot.
The training does require external GPUs (but we provide that infra for free, straight from the app!), but the onboard jetson can run models trained though, as you can see in the examples. Everything you see in the vids is running onboard when it comes to manipulation, because we use a special version of ACT made specifically by us for this robot, that also includes a reward model (like DYNA does).
We have developed this system to also be able to run the other components smoothly so it also does SLAM, and has room for more processing even when running our ACT.
Now indeed this cannot run Pi-0 but from our experience - and the whole community in general - VLAs are not particularly better than ACT in the low data regime, and need a lot more compute.
As for community-level datasets, yes this is the plan. Anything you train can already be shared with others - just share the files. We didn't develop a centralized place for sharing datasets and behaviors but it is on the plan.
Hi, I am currently considering a Lekiwi build but I am intrigued by Mars. Outside of the need for external compute, what issues did you find with SO101 and Kiwi?
Also I am curious about a couple of the parts, if you don't mind sharing - are those wheels the direct drive wheels from waveshare? And what is the RGBD camera? (Fwiw, even if it's hefty the MARS price tag seems fair to me).
and much higher interfaces for interaction and ai manipulation. Like directly recording episodes of training data so that the arm can use a VLA instead of simple IK.
Had a chance to see a live demo last month. Looks great, can do a lot more things than an SO-101 and the teleop via the app if both fun and useful. Would definitely buy if I had the money.
Ok I ordered one. I think yall have something interesting here for sure.
I especially like that you’re using ACT and BC to bootstrap be authoring process. Hopefully behaviors are modular and transportable - which I assume they will be given then arch.
We spent a lot of sweat to make it all transferable from one robot to another easily. It is our goal that we all have robots that feel alive and that we can all develop together.
That's a cool little robot, and I see the appeal, definitely as an educational toy. I don't think the fidelity is there for real autonomous research. But the bigger issue imo is that there's no way this should cost $2000.
You've got a $250 computer, some lidar+camera sensor for maybe $1-200, 6 servos, and cheap plastic. Plus you want to charge a $50/mo software subscription fee for some software product, whatever I guess that's beside the point.
No shade on the idea because low-cost robotics is an unsolved need for the future. But this current iteration is just not competing well with other alternatives. Perhaps this is more of a comment on what we can accomplish in the West vs what's possible in Asia.
Actually the BOM cost required to make something stable that can execute manipulation tasks well enough is around $1k+ hence our price. You will find very cheap robots that can pretend to do what this one can, but in practice won't work well enough.
As for the unitree robot, this one is not unlocked for development, does not have onboard GPU, and does not have an arm. If you want it, check the price they give, it's very prohibitive.
You could attach a cheap arm to it but it would also not be stable enough for AI algorithms to run it. We're researchers ourselves, we would have made it cheaper if we could, but then you just can't do anything with it.
Our platform will deliver the experience of a real AI robot, anything cheaper than that is kind of a lie - or forces you to assemble and calibrate, which we do for you here. It is just the nature of trying to deliver a really complete product that works, and we want to stand for that.
I don't want to be too negative, but all your demos seems extremely silly.
Sure, the package is really interesting and definitely got me interested. But not one of the demos seems like a good use of the hardware. If you want to position yourselves mainly as an educational tool I don't think that is a problem. But if you want to target the 'maker' community I think you should put some thought into that.
For example, you could change the 'security guard' demo into a 'housekeeper' demo. You make it roam your house during the day and keep an up-to-date list of things you need to buy. I think this should work reasonably well for laundry and cleaning products. And after you have some historical data you could even do some forecasts about when you need to buy things again.
Another example would be to have it integrated with weather data and when it starts to rain it goes around the house to check if all windows are properly closed. On this same note it could keep track of the window state during the day and send you a reminder to open/close some windows if the temperature/humidity is above/below some threshold.
I think that by having some more 'useful' examples you should be able to get more attention from the 'maker' community. My guess is that a lot of folks that are heavily into home automation would love to have a device like that help with random things around the house.
Best of luck with your product, and I hope you succeed because this idea looks really exciting.
That's just one example that came to mind. I guarantee I could dig for 30 mins and find a mobile manipulator platform from China that kills it on hardware-to-price ratio that is either 'open enough' or could be made so.
As someone who's dabbled in this before, I guess I'd rather just sit down and plan a BOM and do it myself if that's your markup anyways. Not that it's totally unreasonable for people who just want something super simple out of the box that works.
My general commentary is just that it's sad how much basic servos and what not cost in North America. We've completely ceded this industry to Asia.
Actually the reason is that with the Unitree products if you want the Python SDK the price jumps to $5,000 for the same hardware. At least it was the last time I checked earlier this year.
As a side note, the previous generation of research platforms for that size made in Asia were the Turtlebots, which go for that same price, but without GPU, arm...
I would say the problem is that most manufacturers, including chinese, sell you platforms that are not reliable enough for AI manipulation, and there's a race to the bottom for it, to which we try not to participate to
> I would say the problem is that most manufacturers, including chinese, sell you platforms that are not reliable enough for AI manipulation, and there's a race to the bottom for it, to which we try not to participate
Pretty lofty claims though, really think you're so above everyone on quality at this price point? I know what dynamixels are capable of, and I see the jitter in the demo videos.
Why aren't the manipulator specs easily accessible on the website? Have you run a real repeatability test? Payload even?
It's a neat high-fidelity garage build platform, but I don't see any reason to assume this price premium is due to hardware quality.
The kind of person who would buy this doesn’t care whether it’s $2000 or $5000 probably. They care more about whether it actually exists and will arrive. Complaining about price, especially for something so niche, is useless feedback if you’re not actually in the market for one.
Look fun to me, and you got me hook on the chess roasting robot. Wish I have spare USD2K for the robot hardware.
Just wondering what the main function of the open onboard agentic OS built on top of ROS2? Does it has a dedicated name or just plugin extension for ROS2.
The agentic OS allows to abstract away from ROS2 by creating the concept of "skills" that can be triggered by the agent; you don't really have to care anymore about defining nodes or racing conditions between them, our framework is a lottle simpler in that way. Triggering a skill is the equivalent (almost) to triggering an action in ROS2 btw.
But the interest of it is how it is packaged and that it can be triggered by this VLM-based agent called BASIC that we created.
This agent has a special kind of architecture which gives it spatial memory, capability to react in time, and ability to navigate. This means that skills can be triggered in the right place at the right time, like a true entity. BASIC can interrupt skills in execution depending on how you configure it. So that if it's chilling navigating around and it sees you, it would interrupt navigation and get to you, shake hands and ask if you need help instead of finishing the task without reacting.
This kind of architecture is very similar to what Physical Intelligence used for Pi-0.5 (a VLM triggering a VLA in different areas), albeit to a smaller size for now.
this website froze my computer for several minutes, like i'm browsing the web in 2004. making the impression that the robot will work the same way, very uninterested now
ain't got time for that. happened in both firefox (locked down with lot of privacy settings/adblockers) and chrome (no privacy/adblockers etc). froze entire computer on both browsers for a few minutes for each browser. my computer has like 16gb of RAM
I believe the core challenge in AI robotics is: Can we transfer the cultural knowledge inherent in human bodies and memories?
It's very difficult. Hard to transfer norms, rituals, and intuitive social cues passed organically drives human actions and evolution by enabling adaptive cooperation, empathy, and innovation in diverse societies.
For example, which books to read and whom to trust. You often make decisions on gut feeling which is hard to transfer.
Very cool! But I think we can help you improve your video-streaming and teleop (it seems to be pretty low frame rate in the demo). We've built the probably best remote teleop solution for robotics in the market today and it can be embedded anywhere (white-label). Want to get in touch to discuss? You can find my linkedin in my profile.
the video stream will get a lot of improvement for sure!
the teleop is pretty good already i believe, but happy to chat further. feel free to dm / join discord
Looks awesome!
This isn't so clear though: https://docs.innate.bot/main/software/basic/connecting-to-ba...
> BASIC is accessible for free to all users of Innate robots for 300 cumulative hours - and probably more if you ask us.
Is BASIC used just to create the behaviours or to run them too? It sounds like this is an API you host that turns a behaviour like "pick up socks" into ROS2 motor commands for the robot. Are you open sourcing this too, so anyone can run the (presumably GPU heavy) backend?
Does the robot needs an internet connection to work?
Also, more importantly, what does it look like with googly eyes stuck on?
On BASIC: Yes it does require an internet connection and until we figure out how this works for you it will remain free of use!
It is required to run them, not to create them. And it's not about running "pick_up_socks", this one can already run on your robot. BASIC is required to chain it with other tasks such as navigating to another part of your house and then running another skill to drop the sock somewhere for example
Thank you for the remark, we will make it clearer in the docs
As a consequence: The robot does not necessarily require Internet to run, but if you want it to chain tasks while talking and using memory, yes it does.
As for the googly eyes, give me a minute...
How complex of tasks can you give it? Seems, not to be punny, super basic.
Can it do complex tasks like "pick up socks from room A, drive to room B, and put in basket"? Is the intention to allow hobbyists to do actual work with it or is this version purely novelty rather than a functional "personal robot"?
Additionally, what is the limitation on speed of movement? It seems very slow in movement, is that intentional for safety or is that purely because of running the AI model locally?
The example you gave is exactly this. The driving around in the example use-cases shows an example of going from room to room and execute policies in context:
https://docs.innate.bot/main/welcome/mars-example-use-cases
I take from your comment that the full capabilities of the robot are not properly represented and I take a note to film longer ones. And it can definitely do what you just asked. And multiple times in a row. I will note that it depends on the training quality of course.
On speed of movement, I now realize we didn't mention it anywhere so I added it in the overview but it's pretty fast. 0.7m.s-1 for the base, and the arm can be tormented quite a bit. Just took this video for you:
https://youtu.be/H-gAaTKLm9c
Oh wow, didn't expect this detailed of a response. Thanks!
1 reply →
Coming with a Lidar out of the box seems nice.
Does the MARS hardware really remove the hidden extras (computer with a gpu) mentioned as the downside of HF SO-101 or LeKiwi? While a jetson is good for inference, I feel like to train VLAs you would need access to a powerful GPU regardless. For Lerobot based hardware training ACT was relatively low profile if you use low resolution for the camera feeds, but with increased resolution or with more than one camera I already saw needing more than 8GB of VRAM. If VLA is on the table, finetuning something like the open sourced version of pi0 should already necessitate access to more than one 4090 or above I think.
Also, do you have plans for community-level datasets? I think Lerobot sort of does this with their data recording pipeline and HF integration.
One of our objectives here was to fix everything that we don't like about the SO-101 and the Kiwi, which have several hardware and software flaws in our view. Including, yes, the constant need for a computer to simply run your robot.
The training does require external GPUs (but we provide that infra for free, straight from the app!), but the onboard jetson can run models trained though, as you can see in the examples. Everything you see in the vids is running onboard when it comes to manipulation, because we use a special version of ACT made specifically by us for this robot, that also includes a reward model (like DYNA does).
We have developed this system to also be able to run the other components smoothly so it also does SLAM, and has room for more processing even when running our ACT.
Now indeed this cannot run Pi-0 but from our experience - and the whole community in general - VLAs are not particularly better than ACT in the low data regime, and need a lot more compute.
As for community-level datasets, yes this is the plan. Anything you train can already be shared with others - just share the files. We didn't develop a centralized place for sharing datasets and behaviors but it is on the plan.
If these are intended to be single-dwelling or single-workplace, is there a need to have any onboard processing greater than a few watts?
You could simply host the raw grunt in a base station somewhere else in the premises, keeping the device lighter and lower power.
3 replies →
Hi, I am currently considering a Lekiwi build but I am intrigued by Mars. Outside of the need for external compute, what issues did you find with SO101 and Kiwi?
Also I am curious about a couple of the parts, if you don't mind sharing - are those wheels the direct drive wheels from waveshare? And what is the RGBD camera? (Fwiw, even if it's hefty the MARS price tag seems fair to me).
2 replies →
What motors do you use for the arm and what interfaces do you provide (position, velocity, effort)? How long does the battery last when idling?
On the battery, at least 5 hours when idle, and 3 when moving at least
On the motors, these are dynamixels from robotis, and we provide all three of position, velocity, and effort on the low-level SDK (in ROS2 too)
we also provide interfaces for IK
and much higher interfaces for interaction and ai manipulation. Like directly recording episodes of training data so that the arm can use a VLA instead of simple IK.
Had a chance to see a live demo last month. Looks great, can do a lot more things than an SO-101 and the teleop via the app if both fun and useful. Would definitely buy if I had the money.
emoji-salute sir
Ok I ordered one. I think yall have something interesting here for sure.
I especially like that you’re using ACT and BC to bootstrap be authoring process. Hopefully behaviors are modular and transportable - which I assume they will be given then arch.
That is the correct approach in my opinion.
We spent a lot of sweat to make it all transferable from one robot to another easily. It is our goal that we all have robots that feel alive and that we can all develop together.
Happy to have you onboard!
That's a cool little robot, and I see the appeal, definitely as an educational toy. I don't think the fidelity is there for real autonomous research. But the bigger issue imo is that there's no way this should cost $2000.
You've got a $250 computer, some lidar+camera sensor for maybe $1-200, 6 servos, and cheap plastic. Plus you want to charge a $50/mo software subscription fee for some software product, whatever I guess that's beside the point.
No shade on the idea because low-cost robotics is an unsolved need for the future. But this current iteration is just not competing well with other alternatives. Perhaps this is more of a comment on what we can accomplish in the West vs what's possible in Asia.
Why would I not go for this guy for $1600, and attach an arm? https://www.unitree.com/go2
It's not an apples-to-apples product comparison, but you get the point. There's just so much more raw value offered per dollar elsewhere.
Actually the BOM cost required to make something stable that can execute manipulation tasks well enough is around $1k+ hence our price. You will find very cheap robots that can pretend to do what this one can, but in practice won't work well enough.
As for the unitree robot, this one is not unlocked for development, does not have onboard GPU, and does not have an arm. If you want it, check the price they give, it's very prohibitive.
You could attach a cheap arm to it but it would also not be stable enough for AI algorithms to run it. We're researchers ourselves, we would have made it cheaper if we could, but then you just can't do anything with it.
Our platform will deliver the experience of a real AI robot, anything cheaper than that is kind of a lie - or forces you to assemble and calibrate, which we do for you here. It is just the nature of trying to deliver a really complete product that works, and we want to stand for that.
EDIT: You can take a look at our autonomous demos there, you need something reliable for these: https://docs.innate.bot/welcome/mars-example-use-cases
I don't want to be too negative, but all your demos seems extremely silly.
Sure, the package is really interesting and definitely got me interested. But not one of the demos seems like a good use of the hardware. If you want to position yourselves mainly as an educational tool I don't think that is a problem. But if you want to target the 'maker' community I think you should put some thought into that.
For example, you could change the 'security guard' demo into a 'housekeeper' demo. You make it roam your house during the day and keep an up-to-date list of things you need to buy. I think this should work reasonably well for laundry and cleaning products. And after you have some historical data you could even do some forecasts about when you need to buy things again.
Another example would be to have it integrated with weather data and when it starts to rain it goes around the house to check if all windows are properly closed. On this same note it could keep track of the window state during the day and send you a reminder to open/close some windows if the temperature/humidity is above/below some threshold.
I think that by having some more 'useful' examples you should be able to get more attention from the 'maker' community. My guess is that a lot of folks that are heavily into home automation would love to have a device like that help with random things around the house.
Best of luck with your product, and I hope you succeed because this idea looks really exciting.
1 reply →
That's just one example that came to mind. I guarantee I could dig for 30 mins and find a mobile manipulator platform from China that kills it on hardware-to-price ratio that is either 'open enough' or could be made so.
As someone who's dabbled in this before, I guess I'd rather just sit down and plan a BOM and do it myself if that's your markup anyways. Not that it's totally unreasonable for people who just want something super simple out of the box that works.
My general commentary is just that it's sad how much basic servos and what not cost in North America. We've completely ceded this industry to Asia.
1 reply →
Actually the reason is that with the Unitree products if you want the Python SDK the price jumps to $5,000 for the same hardware. At least it was the last time I checked earlier this year.
As a side note, the previous generation of research platforms for that size made in Asia were the Turtlebots, which go for that same price, but without GPU, arm...
I would say the problem is that most manufacturers, including chinese, sell you platforms that are not reliable enough for AI manipulation, and there's a race to the bottom for it, to which we try not to participate to
> I would say the problem is that most manufacturers, including chinese, sell you platforms that are not reliable enough for AI manipulation, and there's a race to the bottom for it, to which we try not to participate
Pretty lofty claims though, really think you're so above everyone on quality at this price point? I know what dynamixels are capable of, and I see the jitter in the demo videos.
Why aren't the manipulator specs easily accessible on the website? Have you run a real repeatability test? Payload even?
It's a neat high-fidelity garage build platform, but I don't see any reason to assume this price premium is due to hardware quality.
5 replies →
>Why would I not go for this guy for $1600, and attach an arm? https://www.unitree.com/go2
Quasi-Lego-style robo dog for RPi is $100-150 on AMZN
The kind of person who would buy this doesn’t care whether it’s $2000 or $5000 probably. They care more about whether it actually exists and will arrive. Complaining about price, especially for something so niche, is useless feedback if you’re not actually in the market for one.
Look fun to me, and you got me hook on the chess roasting robot. Wish I have spare USD2K for the robot hardware.
Just wondering what the main function of the open onboard agentic OS built on top of ROS2? Does it has a dedicated name or just plugin extension for ROS2.
The agentic OS allows to abstract away from ROS2 by creating the concept of "skills" that can be triggered by the agent; you don't really have to care anymore about defining nodes or racing conditions between them, our framework is a lottle simpler in that way. Triggering a skill is the equivalent (almost) to triggering an action in ROS2 btw.
But the interest of it is how it is packaged and that it can be triggered by this VLM-based agent called BASIC that we created.
This agent has a special kind of architecture which gives it spatial memory, capability to react in time, and ability to navigate. This means that skills can be triggered in the right place at the right time, like a true entity. BASIC can interrupt skills in execution depending on how you configure it. So that if it's chilling navigating around and it sees you, it would interrupt navigation and get to you, shake hands and ask if you need help instead of finishing the task without reacting.
Hope this helps!
I'll buy one...but your discount code doesn't work. It says "Enter a valid discount code".
Fixed it! It was only possible to use a given number of times
Can it make sone complex tasks ?
What would you define as "complex"?
This kind of architecture is very similar to what Physical Intelligence used for Pi-0.5 (a VLM triggering a VLA in different areas), albeit to a smaller size for now.
You can also see some example (autonomous!) use-cases here: https://docs.innate.bot/welcome/mars-example-use-cases
> our novel SDK called BASIC
Sigh.
yeah it's a tribute
Doesn’t that name make it confusing for your users? A tribute maybe shouldn’t be identical?
1 reply →
this website froze my computer for several minutes, like i'm browsing the web in 2004. making the impression that the robot will work the same way, very uninterested now
you're the first person to mention that to me and it's very very helpful actually. can i contact you to see what you see?
ain't got time for that. happened in both firefox (locked down with lot of privacy settings/adblockers) and chrome (no privacy/adblockers etc). froze entire computer on both browsers for a few minutes for each browser. my computer has like 16gb of RAM
6 replies →
> A Desk Personal Productivity companion that hits you whenever you look at your phone or when it sees you are not working on your screen.
So now employees will not only surveil employees but also physically punish them remotely if their gaze veers off screen?
What a time to be alive!
I believe the core challenge in AI robotics is: Can we transfer the cultural knowledge inherent in human bodies and memories?
It's very difficult. Hard to transfer norms, rituals, and intuitive social cues passed organically drives human actions and evolution by enabling adaptive cooperation, empathy, and innovation in diverse societies.
For example, which books to read and whom to trust. You often make decisions on gut feeling which is hard to transfer.
The product looks promising. Hoping for the best.
Very cool! But I think we can help you improve your video-streaming and teleop (it seems to be pretty low frame rate in the demo). We've built the probably best remote teleop solution for robotics in the market today and it can be embedded anywhere (white-label). Want to get in touch to discuss? You can find my linkedin in my profile.
the video stream will get a lot of improvement for sure! the teleop is pretty good already i believe, but happy to chat further. feel free to dm / join discord
Don't underestimate the challenges of making remote-teleop work reliably and efficiently with low-latency via the Internet: https://transitiverobotics.com/blog/streaming-video-from-rob....
4 replies →