Welding Seam Scan Pose Planning for Large Object Assembly
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional robotic assembly systems face inefficiencies in scanning large objects due to the generation of massive, potentially useless scan data, slow processing times, and inability to accurately locate seams for welding operations, especially on objects exceeding cubic meter volumes.
Innovation Solution
A robotic system that uses a controller to select optimal scan poses based on candidate poses evaluated for informational content and ease of assembly, employing discrete or continuous scan operations to rapidly and accurately locate seams, thereby reducing unnecessary data generation and processing time.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a brute force scanning approach is used to scan the entire surface of large objects, then complete coverage of the object surface is achieved, but massive quantities of useless scan data are generated and processing time increases significantly
Solution Approach 1:
The patent divides the object surface into multiple regions of interest (ROIs) based on expected seam locations, rather than scanning the entire surface uniformly. The controller segments the scanning task by identifying specific areas where seams are likely to occur and focuses scan operations only on these segmented regions, thereby reducing overall data volume and processing time while maintaining seam detection accuracy.
Solution Approach 2:
The patent applies different scanning strategies to different regions of the object surface. High-resolution scanning is applied only to regions where seams are expected to be located, while other regions receive minimal or no scanning. This local differentiation of scan quality ensures that computational resources are concentrated on areas that matter for seam detection, reducing waste of processing time on irrelevant areas.
2Reliability
If repetitive scanning from the same pose is performed to ensure complete coverage, then all surface areas are captured, but the same pose may not capture useful features and computational resources are wasted
Solution Approach 1:
The patent performs preliminary actions by using representation data (such as CAD models or point cloud data) to pre-identify regions of interest and expected seam locations before actual scanning begins. This preliminary analysis allows the system to plan scan poses and trajectories that will capture useful features, avoiding the need to perform exhaustive repetitive scanning from the same pose and thereby reducing computational resource consumption while maintaining reliable seam detection.
Solution Approach 2:
The patent incorporates feedback mechanisms where scan results from initial poses are analyzed to determine whether useful seam information has been captured. Based on this feedback, the controller dynamically adjusts subsequent scan poses and trajectories, avoiding redundant scanning from unproductive poses and directing computational resources toward poses that are more likely to yield useful seam detection data.
3Measurement precision
If the entire object surface is scanned to locate seams, then comprehensive data is collected, but the approach is infeasible for objects exceeding cubic meter in volume
Solution Approach 1:
The patent extracts and focuses only on the essential information needed for seam detection by using representation data to identify and extract regions of interest from the complete object model. Instead of scanning the entire cubic-meter-scale object surface, the system extracts and scans only the specific regions where seams are expected to be located, making the scanning process feasible for large objects while maintaining the precision needed for accurate seam location.
Solution Approach 2:
The patent transitions from a three-dimensional exhaustive surface scanning approach to a more efficient strategy by using representation data (such as CAD models) to virtually pre-locate seams in the digital domain before physical scanning. This dimensionality shift allows the system to plan scan trajectories that target specific regions, reducing the complexity of physically scanning entire large-object surfaces while maintaining seam location precision.
Data Source
AI summary
Disclosed are systems, methods, and apparatuses, including computer programs encoded on computer storage media, for operation of an assembly robotic system. In one aspect, the assembly robotic system performs at least one of a first or second scan operation. In the first scan operation, one or more scan poses is selected from among a plurality of generated candidate poses. For each scan pose of the one or more scan poses, the controller initiates a scan operation associated with a region identified to include a seam associated with a feature of the object. As part of the second scan operation, for each candidate scan pose, a scan operation is simulated. Based on the generated simulated scan data, multiple scan poses are selected and a scan trajectory is generated for a scan operation. Other aspects and features are also claimed and described.