Artificial intelligence based robot cable simulation

A computer-implemented method for creating a training dataset for training a cable position artificial intelligence (AI) algorithm configured for simulating a motion of a cable of a real kinematic system having at least one joint, which comprises the steps of:

• receiving a virtual representation of the real kinematic system;
• receiving a starting pose and a set of end poses for the real kinematic system, the end poses defining different positions that have to be reached by the real kinematic system;
• simulating, for each end pose of the set of end poses, a motion of the real kinematic system from the starting pose to the end pose, wherein a simulation is configured for simulating the motion of the real kinematic system and of the cable, wherein, at an end of each simulation time interval, performing a calculation of a next joint value from a previous joint value and in a corresponding simulated cable position for the next joint value, a joint value data subset containing the next joint value is collected for populating an input training dataset, and a cable information data subset containing a corresponding simulated cable position is associated to the joint value data subset and collected for populating an output training dataset;
• and
• storing the input training dataset and the output training dataset.

A computer-implemented method configured for providing a trained cable position artificial intelligence (AI) algorithm configured for simulating a motion of a cable of a real kinematic system having at least one joint, which comprises the steps of:

• receiving a training dataset with a first interface, the training dataset containing an input training dataset and an output training dataset, wherein:
• the input training dataset contains several joint value data subsets, wherein each of the joint value data subsets contains a next joint value;
• the output training dataset contains for each of the joint value data subsets, a cable information data subset having a corresponding simulated cable position;
• training the cable position Al algorithm, wherein the joint value data subsets are used as input training data for the cable position Al algorithm and the cable information data subsets are used as output training data;
• and
• providing a trained cable position AI algorithm with a second interface.

A data processing system, comprising:

• a processor;
• and
• an accessible memory connected to said processor, the data processing system configured to:
• receive a virtual representation of a real kinematic system;
• receive a target task to be performed by the real kinematic system;
• receive a starting pose of the real kinematic system;
• use the virtual representation of the real kinematic system for simulating, from the starting pose, a motion of the real kinematic system enabling the real kinematic system to perform the target task, wherein the real kinematic system is simulated without a cable, and wherein a simulation is configured for calculating a next joint value from a previous joint value;
• calculate, at an end of each simulation time interval, the next joint value from the previous joint value;
• use the next joint value as an input to a cable position artificial intelligence (AI) algorithm trained for outputting a cable position for the next joint value;
• store an outputted cable position for the next joint value;
• and
• display the virtual representation of the real kinematic system characterized by the next joint value and its associated said outputted cable position.

A non-transitory computer-readable medium encoded with executable instructions that, when executed, cause at least one data processing systems to:

• receive a virtual representation of a real kinematic system;
• receive a target task to be performed by the real kinematic system;
• receive a starting pose of the real kinematic system;
• use the virtual representation of the real kinematic system for simulating, from the starting pose, a motion of the real kinematic system enabling the real kinematic system to perform the target task, wherein the real kinematic system is simulated without a cable, and wherein a simulation is configured for calculating a next joint value from a previous joint value;
• use, at an end of each simulation time interval resulting in a calculation of the next joint value from the previous joint value, the next joint value as input to a cable position artificial intelligence algorithm trained for outputting a cable position for the next joint value;
• and
• storing an outputted cable position for the next joint value, and to display the virtual representation of the real kinematic system characterized by the next joint value and its associated said outputted cable position.