Open Source Robotics + Robo Dog Demo - Berkeley 2024

[Music] awesome so hello everyone my name is Vitali and I'm robotics sard I'm really into robotics since I was a kid uh and for the past 10 years I've been working with robots with everything from drones Rovers autonomous boats industrial arms and the quadruped robots as well yeah and today I'm going to talk about open source developments in robotics because uh there is actually a big ecosystem of Open Source um robotics tools uh I start by saying and by talking a little bit about the history of Open Source and Robotics and explain what tools exist and then I'm going to offer my perspective and propositions as a web free community what we can add to robotics community so I want to start by saying that it's prime time for robotics right now um the market for service robotics has grown 48% last year and New Market verticals are um just flourishing a lot of new projects are starting um and uh um obviously AI Innovations unlock um a lot of new possibilities in robotics uh we started with industrial drones uh robots like 30 years ago uh but more recently we start seeing robots that navigate our complex world with convolution neural networks before and now with Transformer models uh we see companies training more than 100 tasks each day uh so the unities to implement robotics in new uh be never before possible use cases is now essentially and I want to ask you how many of you have heard of Ross Robot Operating System awesome some some did but not that many so Ross is the dominant open source Frameworks for robotics developers um it has around 150 active developers uh last year and a few million people has worked with Ross over the years um and it has packages for most and the tools for most uh things that robotics developer may need uh for navigating and for planning path of control for industrial arms gazebo is a great simulation tool that would allow you to try things out before you actually deploy it on the physical Hardware so Ross is like IOS and Android combined in a lot of other tools as well in robotics so every single roboticist I know at least at some point have touched Ross some people people don't like it some do but nobody would deny that it is a dominant framework in robotics and it was started in uh essentially out of research first but then quickly got into incubation space called Willow garage uh where uh all the early Innovations in service robotics happened a lot of startups that were successful in robotics over the past 10 years has started in Willow garage and Robot Operating System was developed there as well uh and initially it was more of a researchy project it was was not thought of as a uh something ready for Enterprise adoption for example but then around 2018 um Enterprise companies have starting to look into Ross more AWS released the robo maker which allowed to run simulations of Ross on their Cloud for any scale and around that time we saw much more adoption and in 2022 Google through their company intrinsic has bought um significant portion of Ross two-thirds of developers have uh are now working closely with Google and just a few weeks ago a new structure was announced open source robotics Alliance uh that would be continuing to maintain open source Ross so now is the uh actually a perfect time to get involved with Ross Community because it's uh changing a lot and it requires more contributors and more participation from the uh from the community and so yeah as I said some of the biggest robotic startups were came from will garage and were acquired um have some some of the largest fleets but I think now we're at the stage where um new AI models um can create more than hundred billion dollar Innovations in robotics industry and my assumption is that Ross 2 specifically would probably be at the foundational layer of uh that value creation so yeah Ross is a great framework but what can we do as a community to help them grow and what like learning we have in our um ecosystem that we can offer about assists um there is also one important shift that is happening right now there is a transition from Ross version one to Ross version two and it's going for few years already because it's um um often times difficult projects in robotics so teams take time to migrate to a new uh system but still uh more than half of the developers already use Ross 2 and among many things it includes the robot security it adds real-time control which is very important for more practical use cases but what I wanted to talk today is uh the modular communication stack that uh exists in Ross 2 that didn't exist before uh which means basically that we can change the whole communication stack that exists within Ross uh to anything else um and I think that existing Alternatives that uh that we have now in Ross community are mostly focused for um smaller Fleet operations so primarily uh it's focused on one company operating their Fleet um setting up private networks and just only having authorized connection to the robots I think that communication uh stack of web 3 projects uh can unlock Global accessibility for robotics with the same level of protection as if they're running on private networks essentially um and uh my team has been actively contributing to Ross Community for a while now uh as a part of robonic team we have given a grant to write rust implementation of Ross so that um Ross can be run on more low powered power low efficiency Hardware um the team at Marb has developed an agent that is currently running on the robot and um that allows to set up communication uh over Le P2P between the all the other agents and and uh our platform as well and we already uploaded more than 10 terabyt of data to filecoin network because we believe that uh web free tools are uh very applicable to robotics where operations are usually geographically distributed as well and having a data storage location close to where operations happens makes a ton of sense P2P already built in we just added support for that um yeah I think I may need to um and the idea is that um I'm now running um an a version of their agent on my laptop and there is a version of the agent running on their drow bit and uh no matter what the config configuration of the network will be whether we're in the same local network or um we use different protocols uh we can still continue um uh have stable communication hopefully the uh message was sent now through Le P2P so I have the agent running locally um on my machine and the device is uh running the version of the agent as well and um we can set up any kind of control for the robot and uh right now um the agent that we're running enables communication between my device and the robot it's itself it also Aggregates all the data that is captured by the device all the videos and all the logs from the machine uh and encrypts it with the key of uh my that is stored in my agent on the laptop as well so that uh all the data as it goes back and forth no matter what configuration of the network will be whether we have a server in some remote location and a fleet of drones flying around it completely offline with no internet or we have a setup where uh like this node is connected to the internet and then uploads information to the cloud uh all those scenarios are possible essentially and uh so right now this is just early demo of how lip Tob could be uh used for um robotic deployments but we think that there is an opportunity to be much deeper into Ross uh and have that as an alternative to existing DDS solutions that exist in Ross and that is not a replacement for what exists today I I have to say uh I I have to stress that it is an alternative that is has its unique use cases so primarily it's Global accessibility of the robot whenever you bring your new robot to some location which is remote uh you don't have to set up private Network you don't have to do any new setup for it as long as it's connected somehow to the internet uh you it can get discovered by any node anywhere in the world and uh um also there is no absolutely no uh setup that is required for heterogeneous devices so it could be a drone it could be a Rover it could be any type of device and they could talk to each other as well uh and plus by using l2b for example uh we do get access to uh additional tools that are built by the community as well uh so we do use ipfs for example for U protecting the data uh as it moves through uh the communication stack that we offer and some other web three tools that um uh eventually we want to create a fully working robotics economy where authentication can happen through the public uh key cryptography that's used in blockchain networks storage networks obviously uh can can be used uh in decentralized way um even financing of Robotics and uh payment to robots directly for services could be set up through decentralized infrastructure of blockchain networks as well so I think by making this first step towards implementing ing Communications stch we can unlock opportunities to implement more web three Tools in robotics in general uh in the future So yeah thank you so much