A Full-stack Platform for Robotic Applications (blog post)

[chfritz]

Implementing new robotic applications and operating a large fleet of robots requires a lot of software. ROS is great for developing the software required on the robot itself, but there is so much more that is needed, and that software just doesn’t yet exist off-the-shelf or open-source. A lot of that “other” software spans multiple systems and categories including on-prem servers, cloud, and user-interfaces, and includes things like health monitoring, remote teleop, configuration management, and elevator integration. See the image below for a longer list of examples. In talking with at least half a dozen robotic startups we have learned that many of us have been developing the same kind of software capabilities in isolation, which is wasteful when most of that software does not provide differentiation in the marketplace. We feel that there is a gap, and we propose to fill it. If you are interested, please read more in this blog post we just published:

A Full-stack Platform for Robotic Applications

We’d love to hear what you think, and we especially want to hear from you if you are interested in working with us on this or have a need for one of these capabilities. Please reach out or comment. Thanks.

[chfritz]

Current state of web and mobile monitoring and teleoperation

Howdy - this is a great writeup Christian. There’s only one piece I don’t agree with… …and as a result, the above capabilities do not exist as commercial-off-the-shelf or open-source modules that robotics companies can use. In fact, we (Formant) have almost all of what you’ve talked about here already generalized and run in production with many thousands of systems and hundreds of different morphologies around the world. I think you may have actually worked with our head of engineering (Abra…

(@jam, responding here, since your comment was about this post rather than the current state of web and mobile monitoring and teleoperation per se.)

Yes, of course I know Formant and the other, existing platforms like it and I’d also include the Robot Middleware Framework (RMF) in that list. I think you are right, though, we are slicing the problem differently and as a result have a different focus. Like you said, you seem to be “focusing exclusively on the data & remote operations infrastructure”, where as we propose to develop an open platform for applications. Applications, of course, involve a lot more than data and remote operation. You may indeed have most of the things we list under DevOps/Fleet Management and perhaps also some of the things under Administration, but that still leaves the other five categories which I think are currently under-served. More importantly though, I didn’t get the sense that any of the existing platforms allowed their users (i.e., robotics companies) to develop on top of them. They all seem to have APIs for interoperability, yes, but none of them seem to enable the development of entirely new robotic applications and capabilities per se, let alone sharing those capabilities with others. Please correct me if I’m wrong.

RMF seems like a step in the right direction and, of course, it is open. But as the name says, it seems to be focused on the middle-layer of the stack and doesn’t provide any means for creating UI components that are ready to be used by end-users. Nevertheless, it would probably be wise for the platform we propose to allow integration with RMF at some level.

[jam]

I would love to see what you’re talking about develop Christian. I think that in fact many of the pieces already exist in the open source world, but there’s an incredible amount of very high value packaging work to do here.

In our case - Formant (and probably others, though I’m not too close) is pretty much comprehensive on all but your “Navigation & Motion”, “Localization & Mapping”, “Computer Vision & AI” buckets. The former are generally open source or developed by our customers, and the latter is done at the edge or via data pipeline integrations with either third party services or our customers’ in-house services. All three of those buckets tend to be quite heavily customized (or just fully proprietary) in the production systems we see though.

We do in fact allow for externally-embeddable UI elements and internally-embeddable custom applications, and do manage some software at the edge. It’s true that we haven’t begun to share these custom applications between organizations in a productized way, though - mostly because there have so far been few actual opportunities there. DroneDeploy’s app market is doing a great job here for a very targeted use case, but I think this is a little ahead of the production autonomous robotics community at the moment.

Ultimately our objective is to be fully complimentary to the ROS ecosystem, and to pick up where it leaves off (hence the focus on auto-scaling cloud infrastructure, pretty mobile UIs, integrations with other commercial products, etc). I think you’re right to consider slicing it in a different dimension though.

The RMF work is very exciting, and I’m very much looking forward to seeing it in production systems soon. I’m expecting that some of this work will significantly level up the starting point for new companies that decide to use it, and that your vision of a more “app store”-like model could do the same.