Linking Simulation to the Real World for Robot AI скачать в хорошем качестве

Linking Simulation to the Real World for Robot AI

For simulation to be viable for robotics, especially for creating AI-generated robot behavior, it is critical not only that robot behaviors created in simulation can be seamlessly translated to the real world, but also the data from the real world can get back to that simulation. This allows for closed-loop learning of robot behaviors in simulation, using information about how the system actually performs. Importantly, the data coming back from the real world can't just be data about the robot, but everything else in the robot's work cell, and its environment. In this latest video, we show how READY's ForgeOS can provide this closed loop, by enabling programming in simulation, seamless transfer to the real world, and providing a path for data to be sent back to simulation. ForgeOS not only surfaces the robot motion back to simulation but also the state of all of the tooling, as shown by the tool changer's behavior being accurately represented when mirroring the real system. The ability to surface the traditionally siloed data in a robotic cell in the factory is something that ForgeOS does out of the box, without any additional IoT devices, and it ties directly back to NVIDIA's Isaac Sim.