I personally have more questions than answers (which might be a good thing in a keynote ). The Flowstate Environment seems nice, we will know more when we get into the beta.
For me the underlying platform is more interesting, about which we did not receive too much information:
Is it based on ROS at all?
The one screen showing some code (Intrinsic Product Keynote - May 15, 2023 - YouTube) had no ROS-specific dependencies, but it showed that data seems to be exchanged using Protocol Buffers (which is not very surprising in a Google company), used Bazel for building and the end result was a Python-based container image.
Kubernetes was mentioned in the keynote. Is it used only on the cloud / web-app side, or also “at the edge / on the robot” - similar to the ROS on Kubernetes initiative: Topics tagged kubernetes
Will this underlying platform be open-sourced in some form or will it be closed-source?
Thanks for opening this topic, Gergely. I had been thinking of doing the same, given the new importance of Intrinsic for ROS given their OSRC acquisition this seems like a worthy discussion.
Unfortunately I don’t have any answers just more questions:
Given that Intrinsic is focused on work-cell automation and given that there was no mention of mobile robots at all, what will future development on ROS (at Intrinsic) look like? Will they be mostly on the ROS-Industrial side?
“Is it based on ROS at all?” – and which one, ROS 1 or ROS 2?
Yes, I definitely share the feeling of having more questions than answers.
@kisg, good catch with the time code But my understanding is that the major work on this product must have happened before the open robotics people joined. So to see the impact and interaction with ROS we may have to wait some more.
I would say it will be almost certainly ROS2. Everything else makes no sense at all.
They target their product specifically on software developers rather than robotic developers. This is a good call, given how many people of each group exist. This means on the other hand that the system must abstract away the real hardware very efficiently and must ensure to be compatible to different robotic hardware. That is literally one of the hardest problem in industrial robotics. So, good on them if they managed to solve this.
They mention to be using behavior trees, while all the GUI that has been shown looked more like some sort of state machine with parallelism.
It sounded at some point like they also want to enable third parties to sell what they call skills on their platform. This is of cause a good move and could, if successful, replicate something like the effect of the android appstore. On the other hand, this may also limit the future options to open source robotic software. Because if your company wants to sell a skill on intrinsics platform, it is unlikely to open source the software they use for this skill.
Sorry, this turned out a bit lengthy, but I guess I had some things to get off my chest.
Many pain points in Robotics are not about the front-end or graphical user interfaces. From the keynote, I understand they go for a low-code platform to make it accessible to more people. I believe this is a strategic mistake.
They seemed to be inspired by the productivity increase for web or cloud service developers. But the core tooling in those domains is all text-based. Now we write our documentation in markdown and draw our diagrams as code. This is why they took off in the last decade. So I would prefer to see YAML files and JSON schemas instead of dropdown buttons. GUIs are red herring for complex software development.
But there are some interesting things: the real-time control framework as they mentioned. I have applied for their private beta, mainly to test what they mean by that. I hope they touch on more core issues and the plumbing of robotics as they mention in the keynote.
We’re excited that we’ve been able to share this early look at Intrinsic’s first product and that now people can understand more about what we’re doing. In addition to the video linked above there’s new information on our blog. We’ll be revealing more as time goes on.
@ct2034 is quite right that this is a work many years in the making. Integrating large systems doesn’t happen overnight and we’d rather do things well than rush them out prematurely. We’re definitely going to be integrating with ROS 2 (vs. ROS 1). We don’t want to have to do a migration in a couple years time, and there are a lot of features in ROS 2 that make it better suited for building our product. We are using containers throughout the system – both in the cloud as well as “at the edge”.
I’m sorry we can’t share even more about the Intrinsic platform and product roadmap at the moment, but we definitely hope to see the ROS developer community as collaborators (rather than customers!). We have open source components throughout the system and remain committed to contributing back to the community. Be sure to check out the Iron Irwini release coming out next week, and we’ll have lots more to share in the future.
). The Flowstate Environment seems nice, we will know more when we get into the beta.
But my understanding is that the major work on this product must have happened before the open robotics people joined. So to see the impact and interaction with ROS we may have to wait some more.