With a payload of 100 kilograms, Magni aims to make it easy to prototype a useful mobile robot
There are any number of robotics development platforms out there, and we’ve written about most of them—TurtleBots, iRobot Creates, and more recently robots like Misty. Generally, these platforms are intended to be used for experimenting with sensors and software, or for more socially-oriented applications that don’t involve much in the way of lifting or moving stuff.
A Silicon Valley startup called Ubiquity Robotics believes that there’s an opportunity here, and they’re crowdfunding a robot called Magni that’s specifically designed to handle large payloads for long durations. It comes with sensing and computing out of the box, and Ubiquity hopes it’ll enable hobbyists to create a new generation of practical robotic solutions.
Here’s what you get with Magni:
- Payload: 100 kg
- Drive System: 2 x 200 W hub motors, 2 m/s top speed
- Power: 7 A+ 5 V and 7 A+ 12 V DC power
- Computer: Quad-core ARM A9 – Raspberry Pi3
- Software: Ubuntu 16.04, ROS Kinetic
- Camera: Single upward facing
- Navigation: Ceiling fiducial based navigation
- Battery life: With 10 Ah batteries, 8 hours of normal operation. Up to 32 Ah lead acid batteries can be installed, which will provide 24 hours+ of normal operation
- 3D sensor (optional): 2x time of flight cameras, 120 degree field of view
In addition, Ubiquity is offering Loki, a small and more or less affordable learning platform that you can use to develop applications for Magni.
It’s important to note that Magni (and Loki) are not for novice programmers, and they’re probably not for people who are interested in learning about robotics and ROS. There are tutorials for Magni, but they assume that you have a working knowledge of ROS already. As Ubiquity says, their platforms are for people who have some experience building robots and writing code in ROS, but who have been waiting for a robust and extendable platform that they can afford to experiment with:
We are a group of people who love building robots—some of us have been building robots for decades. We wanted to quickly and easily build ever more awesome robot applications. What we needed was a starting point, a base robot with payload, vision, navigation, compute, and power.
With a base robot like this, we knew we could build our applications in hours, not months. But that robot did not exist! So we have built Magni: it allows you to complete your robot project 2 years faster, $500k cheaper, and without a large multi-disciplinary technical team.
We want to change the game around robotics and empower the guy in the garage who just cannot do it all on his own. Now that person can do it all by himself: with a Magni.
This strikes us as an optimistic vision for a robotic development platform. Not that there’s anything wrong with optimism, but as we mentioned above, people have been making development platforms for years, and they’ve remained a niche product, mostly used by researchers or hobbyists.
Ubiquity believes that there’s a market out there of people who want to do practical things with robots, but who are restricted more by platform availability than anything else. It’s certainly true that most mobile bases that can handle large payloads tend to be on the expensive side, but it remains to be seen whether there’s enough demand out there to sustain even relatively low volume production of a robot like Magni.
Ubiquity is crowdfunding the version of Magni that most people probably want for US $1,000, not including a 3D vision system, meaning that they’ll probably need to sell between one and two hundred Magnis (of varying specs) to reach their fixed goal of $200k.
For more details and perspective on Magni, we spoke with Ubiquity Robotics founder David Crawley via email.
IEEE Spectrum: Why will people want a development platform like Magni?
David Crawley: Almost every startup in Silicon Valley that is doing something with a mobile robot spends about two years and $500,000 developing a mobile robotic base. Typically, they spend this time and money developing capabilities that have been developed before by others in other contexts, namely good motor control, navigation, mobility, power management etc. Our argument is that they shouldn’t do that, and instead buy a platform that is well engineered from us. Once they have put their mobile base together, developing the application is fast.
Having a standard base like this will do two things: It will allow well funded startups to get to market quicker while reducing the cost they currently bear in terms of electrical and mechanical engineering, and it will enable guys in the garage, who can only dream of getting a robot application to market, to build a minimum viable demo themselves. We are actually more excited by the second group. That’s who we want to target, that’s who we think this is for.
Besides the payload, can you describe how Magni is different from a TurtleBot?
We love our TurtleBots, but you can’t really use a souped up vacuum cleaner for many real-world applications.
- The batteries only last maybe 45 minutes on a TurtleBot. Magni’s batteries last all day—our 32 Ah battery option can do normal operation for around 24 hours straight.
- A TurtleBot just doesn’t have the mobility that you need to do things; it will snag on the edge of a carpet, and can’t handle bumps. We designed our robot to work well in any ADA (American Disabilities Act) compliant space and in fact it exceeds almost all the requirements by factors of 3-5.
- Hauling power—our robot can tow a car. This matters, because you just shouldn’t have to worry about whether your platform can haul the load you want.
- Localization. We were disappointed by many of the localization options out there. We developed our own, ultra-low cost localization method that uses SLAM of fiducial markers to unambiguously and robustly determine location.
- We’ve designed the robot for extensibility. This means that we have 7 A 12 V and 7 A 5 V power supplies on board for whatever the user cares to use them for.
- Lastly, we are putting all this functionality into a package that costs less than a TurtleBot 3 Waffle, less than a TurtleBot 2, and less than a TurtleBot Euclid. Our silver model will crowdfund for $999.
In short, Magni can be used for real-world applications while TurtleBots, for the most part, cannot. And we provide that functionality at a price that you can’t replicate on your own.
Many of the sample applications that you show for Magni, like delivery and telepresence, are already available in commercial platforms. How will you convince people to invest the time and effort to get Magni to do something similar?
These applications are not really the point of Magni. Sure, Magni can do them and they make demos that everyone understands, but what we want is for people to develop something new. The point of Ubiquity Robotics is that it is intended to be an organization that develops the hardware that everyone needs, but isn’t specific to any given application. Ubiquity Robotics focuses on foundational capabilities, supply chain, operations, and design, while other businesses should focus on things like value-added software, systems integration, and customer relationships.
What kind of autonomous navigation is Magni capable of?
Magni has a full ROS stack, so you can use the same autonomous navigation capabilities that are available on any ROS robot. We have also developed our own autonomous localization and navigation stack.
We mentioned our fiducial based localization system. We like it because it’s robust. We also have developed our own super simple navigation system—it uses the same bindings as move_base, but we’ve called ours move_basic. Move_base is terrific, but setting it up and getting it to work well for your application is well beyond the capability of most guys in the garage. Move_basic is very very simple, easy to set up, and robust.
When you say, “Magni is easy to use, you just plug it in and go,” can you describe how using or programming Magni is easier than using or programming a TurtleBot?
We are very spiritually aligned with the goals of TurtleBot as an enabler for people for whom ROS is just too complicated and difficult. We are targeting users for whom robotics is just out of reach now. We aren’t focusing on programming novices yet, but rather people who find setting up a ROS robot daunting.
I think both us and TurtleBot have spent a lot of time trying to flatten the learning curve of ROS, because that expands the set of people who can cope with it and by extension develop on it.
Ease of use is really about doing a lot of little things properly. Essentially there are three core capabilities that we have given to the robot:
- Voice recognition: This is actually housed on a smartphone that communicates with the robot;
- Location awareness and localization;
- Object awareness: The ability to detect and determine the location of objects. This is done using a deep neural net (DNN) system.
Our system uses voice recognition and then attempts to parse voice commands into verb+object commands. So for example if one tells the robot “find bottle,” it will initiate a search routine that causes the robot to look around the room for the object “bottle” until the DNN detects a bottle, and then it will indicate the location of the bottle by driving towards it. You can do a similar thing with location awareness.
I wouldn’t want to call this series of commands a program, but it is a way in which someone (potentially with no programming experience) could start to interact with the robot and get it to do things without even touching a keyboard.
What kind of feedback have you gotten from the robotics community, and how has it affected your development process?
At the very start of the project a list of requirements circulated around the HBRC (Home Brew Robotics Club) for what HBRC members wanted to see in a robot. I talked with various other members of HBRC and others who had been playing with robots in their garage (some for 30 years) about what they wanted. The underlying theme of course was, “We love TurtleBot, but you can’t do much real life stuff with it.”
A lot of the time, the gap between vision and reality comes down to what you can actually manufacture. We took a different approach to most people. Generally people describe the robot that they want and then design that, and then try to figure out how to make it. We took the opposite approach, we looked for what was easy to build and then designed our robot around the supply chain rather than the other way around. The result is a robot that gives you a lot more bang for the buck.
I think the most important thing was that we had people on the project who had been doing various robot projects in their garage for decades. We built what we wished we had, and that was always the mantra.