Apparatus and method for controlling a robotic device

Abstract

Apparatus and method for controlling a robot device, wherein the method has: training a control model having a parameter model and an object model, the training having: for each initial state-target state pair of a plurality of initial state-target state pairs, providing a control-state sequence having states and state transitions, wherein each state transition is assigned a set of task parameters; determining a set of state transition-state-state transition triplets, which triplets are contained in the provided control-state sequences; for each state transition-state-state transition triplet: adapting the parameter model such that the parameter model determines a probability distribution for each task parameter of the set of task parameters assigned to the state transition following the state, adapting the object model such that the object model determines a probability distribution for an object state for each object of one or more objects; and controlling the robot device using the control model using the trained parameter model and the trained object model.

Images