Robin graduated and is at Boston Dynamics. 5 years ago, pre the release of Julia v1.0, a lot of the tools for targeting embedded were missing but now lots of the toolchain for that is there (though it's not perfect by any means).
@@chrisrackauckasofficial For my part does it seem like the main bottle neck of Julia is being able to communicate and parallelize between nodes or something equivalent (like in ROS2 software architecture) in a elegant fashion. It makes it possible to run control, detection and mapping algorithms separately while also being able to analyze the message between and maybe mocking nodes. I dont know, but from a julia-standpoint does it seem hard to make this work smoothly because of the need of pre-compilation etc. It is probably possible, but may not be the most optimal robotic software architecture/philosophy in Julia.
What is the smallest robot that can be programmed? If you self replicating robots that replicated the,selves smaller each time you could have a swarm of billions of robots. You could could hop in between the robots to do different states and stuff.
Instead of trying to simulate robot and all of the rigid body dynamics, why not just have random wind and acceleration factors and the robot would predict its path through multi regression and change it si that it goes into a stable configuration
Why Julia rather than Python?
Speed, multi-threaded, you can write GPU kernels in Julia
I see a big red flag when julia-roboticists neglect running their code on actual robots
Robin graduated and is at Boston Dynamics. 5 years ago, pre the release of Julia v1.0, a lot of the tools for targeting embedded were missing but now lots of the toolchain for that is there (though it's not perfect by any means).
@@chrisrackauckasofficial For my part does it seem like the main bottle neck of Julia is being able to communicate and parallelize between nodes or something equivalent (like in ROS2 software architecture) in a elegant fashion. It makes it possible to run control, detection and mapping algorithms separately while also being able to analyze the message between and maybe mocking nodes.
I dont know, but from a julia-standpoint does it seem hard to make this work smoothly because of the need of pre-compilation etc. It is probably possible, but may not be the most optimal robotic software architecture/philosophy in Julia.
What is the smallest robot that can be programmed? If you self replicating robots that replicated the,selves smaller each time you could have a swarm of billions of robots. You could could hop in between the robots to do different states and stuff.
Instead of trying to simulate robot and all of the rigid body dynamics, why not just have random wind and acceleration factors and the robot would predict its path through multi regression and change it si that it goes into a stable configuration