Robot Programming Using Learned Pose Corridors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current robot programming methods are inefficient and inflexible, particularly for tasks requiring tolerance of deviations in object positioning and geometry, leading to high engineering costs and limited automation in industries like automotive, where small series production and flexible components necessitate frequent reprogramming and precise adjustments.
Innovation Solution
A method and system that utilize predefined movement templates with parameterizable execution modules and learning modules to record configurations and calculate parameters using machine learning, allowing for robust and flexible robot programming that tolerates deviations in movement paths, gripper positions, and measured forces, enabling dynamic and variable movement planning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional waypoint-based programming methods are used, then the robot program structure is simple and clear, but the program cannot tolerate deviations in positioning and geometry, leading to program failure
Solution Approach 1:
The patent transforms fixed Cartesian waypoints into flexible pose corridors by changing the parameter representation from precise coordinates to tolerance-based regions. This allows the robot to tolerate deviations in positioning and geometry while maintaining program structure through learned parameters that adapt to real-world variations.
Solution Approach 2:
The patent creates simplified representations (pose corridors) that copy the essential characteristics of complex movement patterns without requiring precise replication of every waypoint. The learning module learns from demonstrated movements and creates parameterized models that capture the intent while tolerating deviations.
2Adaptability or versatility
If dialog-supported and textual programming are used, then complete robot programs with sensor evaluation and force control can be created, but expert knowledge of robot programming language and process is required
Solution Approach 1:
The learning module enables the robot to learn movement patterns autonomously from demonstrated configurations without requiring expert programming knowledge. The system serves itself by automatically generating pose corridors and parameters from teach-in data, eliminating the need for experts to manually program complex force control and sensor evaluation sequences.
Solution Approach 2:
The patent introduces pose corridors as an intermediary representation between simple waypoint programming and complex textual programming. This intermediary layer provides the flexibility of sensor-based programming while maintaining the simplicity of template-based approaches, acting as a mediator that bridges the gap between ease of use and programming power.
3Productivity
If fixed programmed trajectories are used, then automation can be achieved, but frequent reprogramming is required for small series production and flexible components
Solution Approach 1:
The patent transforms static fixed trajectories into dynamic pose corridors that can adapt to different components and positioning variations. The learned parameters in the execution modules allow the same program to handle small series production and flexible components without frequent reprogramming, as the system dynamically adjusts within the learned tolerance regions.
4Ease of manufacture
If teach-in method with manual waypoint saving is used, then the programming process is straightforward, but high engineering effort is required and the method is inefficient for complex paths
Solution Approach 1:
The patent segments the programming process into two phases: a one-time learn phase where configurations are recorded, and an efficient execution phase where the parameterized program runs automatically. This segmentation reduces engineering effort for complex paths by eliminating the need to manually program each waypoint while maintaining straightforward operation through the learned pose corridors.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
The invention relates to a method for programming a robot, in particular a robot comprising a robotic arm, in which method a movement to be performed by the robot is set up preferably in a robot programme by means of a predefined motion template, the motion template is selected from a database comprising a plurality of motion templates, the motion template comprises one or more execution modules that can be parameterized and at least one learning module, the one or more execution modules are used for planning and/or performing the robot movement or part of the robot movement, the leaning module records one or more configurations of the robot during an initialization process, in particular in the form of a teaching process, and the learning module calculates parameters for the one or more execution modules on the basis of the recorded configurations, preferably using an automatic learning process. Also disclosed is a corresponding system for programming a robot.