Robot Simulation#
Note
We are introducing a new Motion Generation (Experimental) API which aims to package all Isaac Sim control algorithms (wheeled robots, manipulators, humanoids) under a single framework. This new API provides a flexible controller composition system and simplifies building collision world models from your USD stage. This extension is documented under the Robot Motion (Experimental) section.
The Robot Simulation section provides information on tools that you will need to move a robot. The lowest level of control is joint control. For the next level up, we separated the controllers by the robot types, for they represent the three types of controllers we provide in Isaac Sim:
Wheeled Robots: use controllers that are based on universal formulas and require very few robot-specific parameters as inputs.
Manipulators: use controllers that are based on complex optimization, therefore the same robot performing the same task could use many variety of controllers, each with a different optimization method. They often require the robot models in the optimization process.
Policy Controlled Robots: uses controllers that are trained using reinforcement learning. They also has a much looser definition “controllers”, for they can have task and path planners embedded as well.