 green cheerful, but that's fine. We will just we'll just skip that part since we used a little bit of time anyway Anyway, as mentioned so yeah starting again Adrian Mcnell co-founder and CEO at Fox Glove and As I mentioned before starting Fox Glove. I spent five years leading infrastructure at cruise and Before that I was also at Coinbase leaving the engineering team, but forget about crypto and NFTs today We are going to talk about something that's far more important than crypto and that is robotics So by the end of this decade robots are going to be everywhere Robots are already really common today in industrial automation But these robots are largely confined to cages and the next generation of intelligent Collaborative robots is going to completely transform the economy in our society. I Spend a long time working in self-driving car space But what I think is even more exciting than self-driving cars is this technology spreading to everything from autonomous tractors Autonomous forklifts self-driving lawn mowers Shipping and logistics robots cyborg delivery underwater robots and and much more For this robotics revolution to happen. There's still a whole missing industry of robotics developer tools and infrastructure so Developing and deploying robots requires a lot of specialized tooling That's things from deployment to fleet management tele operations data management Simulation and visualization And we're starting to see off-the-shelf solutions here, but it's still really early and most companies are still building everything in-house This is different from every other area of Self-ware development and machine learning where there's a huge number of tools and libraries and frameworks to make your life easier So this lock lack of off-the-shelf tooling is slowing down robotics development for academics professionals and hobbyist developers alike Of we're pushing forward open-source robotics infrastructure and developer tools So we created fox club studio and open-source visualization tool and we also we also created mcap Next generation robotics logging library We're maintaining a number of Ross packages and also we have about 20 different types group libraries for interacting with Ross and other robots Today, I'm going to focus on visualization specifically because visualization is a critical part of robotics development Robotics development is not like traditional software. So we can't just look at text logs to understand our robots behavior We need to be able to understand how our robot sees and perceives the world What are my senses seeing? How did the robot build a model of the world? What did the robot plan to do and what did it actually do? Robots generate a lot of data far more than traditional software We have all sorts of sensors and sensor data like cameras point clouds depth clouds GPS IMU wheel encoders, and we also have a lot of internal robot state that we want to record So this is things such as objects that the robot is tracking maps and grids planned actions and controls data and Visualizing all of that data is really important for everyone working at a robotics company From robotics engineers and machine learning engineers to product managers qa teams and operations teams We need visualization to debug live robots or to playback recorded data We can triage failures so we can replay those failures and simulation And we can also compare the results of different experiments in either Either the real world or in simulation Before we can visualize data though We need to get it off the robot and a universal problem in robotics is network bandwidth We almost never have enough bandwidth to stream all of the sensor data off a robot at once So typically we find robots will send lightweight telemetry data And this will give us an overview of the fleet status and the fleet health This can be things like the robot's location Current task it's working on or low resolution camera images And on the other hand The raw high definition log data is usually recorded to a hard drive on the robot These are things like raw camera images Other sensor data detailed recordings of the robot's thoughts and actions They take up a lot of room these can be anywhere from Ends of gigabytes to over a thousand per hour of data and This data can be uploaded when the robot returns to base or sometimes it's just fetched manually by the engineers when needed To view and replay this robotics data there's several options ross ships with a handful of tools such as avers and rqt And if you want to play back recorded data, you need to supplement these with command line tools such as ross bag and ross topic So a pretty typical ross workflow Involves at least three or four different windows open rearranged on your desktop just to play back a single recording And for robots that don't use ross Then there's nothing up until now you either have to use ross or you have to build all of this from scratch But visualization needs are pretty standard across organizations And this type of tooling should be available off the shelf without being coupled to a specific robotics framework the current generation of ross tools have They've served us well over the past decades, but they have several limitations some of these that we ran into at cruise First of all, they're linux desktop apps. So they're very difficult to get working on windows or mac They also require your ross code base to be installed and built locally So that makes it difficult for product managers and qa engineers and operations teams to access that data without a full development environment And the existing tools are only designed for live visualization So as I mentioned playing back recorded data often requires you to open a separate terminal window There's no way to seek and scrub playback Finally, it would be nice if all of these visualizations were integrated into one single tool So that's why we built fox glove studio an open source IDE for robotics So web-based platform So you can view data directly on your browser and download native apps for mac windows and linux We support both ross and non-ross data So you can easily drag and drop a ross bag file or you can use our python libraries to connect to different types of robots Fox glove studio actually began as a project called webbiz that we started at cruise in 2017 and we open sourced in 2019 But webbiz only supported ross one and it had many features that were specific to cruises needs So in 2021 we created fox glove to expand this technology and to bring it to the world Fox glove studios open source We have an active community and github and slack to respond to questions bug reports feature requests And we see a really really wide range of users across the community. So this is hobbyists to academics small companies to large companies Fox glove studios built to be extensible We already support a wide range of data formats. Like I said ross one and ross two We also support other common data formats such as protobuf and json And fox glove comes with about 20 different types of visualizations that you can use out of the box We also have an extension api so that you can build your own Finally, it's collaborative. So you can use it as an individual So like I said, it's open source You can use it as an individual connected to a local robot or viewing files on your computer But you can also sign in and share views and share data across your team Besides fox glove studio for visualization. We also created the mcap format for logging and storing robotics data So this is a pretty new project that we just announced in the past couple months Many companies that we spoke to were unhappy with the existing ross bag format Or that there's a lot of companies that are not using ross at all And so as a result, we found lots of companies that have created their own proprietary file formats in-house for recording robotics data We decided to create an open source format that's specifically optimized for storing robotics data So mcap is pretty similar to ross bag files, but it's not limited to ross and it can be used by any robot We do have an mcap plugin for ross two We also have libraries available in five different languages so you can integrate this with any non-ross robots And you can easily view and open these mcap files in fox glove studio Just as easily as a bag file to replay and visualize them So I was going to jump into a live demo here But since my screen chair is not working and my gif's not working either You'll just have to you'll just have to picture it if you head on over to foxglock.dev There's actually someone's thrown the link in the chat there You can you can take a look at it. We have some videos that will let you play through that So I suggest taking a look you can also just click on view live demo and you can jump straight into the app And and give it a go yourself If you have any bag file sandy just just grab a bag file drag it in that all works locally Even though it's running in the browser. We can just open bag files off your desktop and interact with them directly Happy to jump into any questions. I know we're a little bit behind schedule. So Yeah Any any questions people have feel free to put them in the chat