Abstract: At least one robot is disclosed for object manipulation, on the basis of a probabilistic approach without object detection by visual sensors, a probability distribution is determined approximately by an iterative, recursive application of a Bayes filter algorithm, wherein state transition probabilities obtained, in a nondeterministic motion model for system simulation, are multiplicatively linked with measurement probabilities obtained at the start of the application.

Abstract: In order to be able to automatically eliminate discrepancies, arising in the course of the configuration of a robot-object system environment, between the reality of the robot-object system environment and its digital representation as a CAD model, without manual on-site commissioning of the robot-object system environment with adaptation of the CAD model to the reality, the following is proposed for configuring a robot-object system environment having at least one object and having a robot for object manipulation and object sensing: synchronizing a digital robot twin, which digitally represents the robot-object system environment and controls the robot for the object manipulation on the basis of a control program, for expedient use of the robot in the robot-object system environment during the object manipulation, appropriately and, in this regard, in one or two stages.

Abstract: Provided is a method wherein a movement or a manipulation of an object by a robot, observing constraints, from a starting condition to the manipulation target, is generated, wherein the manipulation is divided into different manipulation modes or sections, which include different constraints, wherein a plurality of manipulation mode-specific controllers for controlling partial manipulations in the different manipulation modes for sections are randomly generated and an optimized sequence of the controllers is randomly generated, wherein the controllers specify a vector field or a directional field, wherein a simulation module simulates the manipulation or movement for each of the controller sequences and determines an expense value or a cost value quantifying a reaching of the manipulation target, and wherein the controller sequence and the expense value are supplied to a machine learning module as training data in order to indicate an expense-optimized controller, which optimizes or minimizes the expense valu

Abstract: At least one robot is disclosed for object manipulation on the basis of a probabilistic approach without object detection by visual sensors, a probability distribution is determined approximately by an iterative, recursive application of a Bayes filter algorithm, wherein state transition probabilities obtained, in a nondeterministic motion model for system simulation, are multiplicatively linked with measurement probabilities obtained at the start of the application.

Abstract: A method for determining an optimized sequence of movements of a robot device for moving a first object in such a way that the first object is brought into a target position independently of an uncertainty of a position of the first object in relation to a second object and/or independently of an uncertainty of a position of the robot device is provided. The method includes: simulating movement portions of the robot device taking account of the uncertainty of the position of the first object and/or the uncertainty of the position of the robot device; and determining the optimized sequence of movements of the robot device taking account of the simulated movement portions and boundary conditions which specify at least one starting position and the target position of the first object. An optimized sequence of movements is determined, by which the robot device can guide the first object reliably into its target position.