To increase robot autonomy it is necessary that a robot can reason over longer time lines than is currently possible with minimal input of a human operator. We propose to develop a system (on top of ROS 2 and MoveIt 2) that enables operators to command humanoids at a high level while still have the robot's motions satisfy task-specific low-level motion constraints. This is done in two steps. First, we will develop several parametrized primitives that allow an operator to specify a broad range of constrained motions, including climbing in microgravity, turning valves, and opening doors. Second, we will develop a task construction framework that allows an operator to compose such primitives into much larger tasks. The task construction framework will be able to compute complete end-to-end continuous motion plans that satisfy all relevant motion constraints. The operator will be able to preview these plans and select one for execution. We will also develop appropriate user interfaces that allow operators to specify complex tasks without having to write any code.
The proposed system would enable operators to specify complex sequences of motions for humanoid robots (such as Robonaut 2) in microgravity environments like the International Space Station or the Gateway. The system will automatically compute feasible paths that can be selected by the operator for execution. Applications of the system include (but are not limited to) planning humanoid climbing motions aboard the ISS or Gateway, opening valves/doors, and retrieving bags from storage.
In warehouse logistics, manipulators often need to complete a series of pick-and-place operations. Such a task can be completely specified using our system by an operator who does not need to be a robotics expert. Besides order fulfillment in warehouses, applications include remote operation of inspection or search and rescue robots. Eventually, assembly or food preparation tasks could be enabled.