Reinforcement Learning for Robot Inspection Imaging Optimization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In flaw inspection using robots, existing methods struggle to optimize the number of imaging points and positions to ensure thorough inspection without overlooking small flaws, while efficiently managing cycle time, due to variations in camera, workpiece, and illumination positional relationships.
Innovation Solution
A machine learning device employing reinforcement learning with a robot control system that adjusts imaging points and positions based on flaw detection information, optimizing the number of imaging pieces and positions by carefully inspecting high-risk areas and briefly inspecting low-risk ones, using a reward system to update action value functions and adjust imaging regions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If the entire range is carefully inspected with increased resolution and small field of view, then measurement precision is improved, but productivity deteriorates due to repeated imaging
Solution Approach 1:
The patent applies local quality by differentiating inspection intensity across different regions of the workpiece. High-risk areas with historical flaw data receive careful inspection with multiple imaging points and varied angles, while low-risk areas receive brief inspection with fewer imaging points. This selective approach maintains high detection precision for critical regions while reducing overall cycle time.
Solution Approach 2:
The patent implements partial action by performing exhaustive inspection only on high-risk areas identified through historical data and performing minimal inspection on low-risk areas. The system determines the minimum necessary imaging points and angles based on flaw probability, avoiding excessive imaging in regions where flaws are unlikely to occur.
2Measurement precision
If multiple imaging positions are used to consider positional relationships, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent applies dynamics by making the inspection plan adaptive rather than static. The system dynamically determines the number and positions of imaging points based on historical flaw data, workpiece characteristics, and risk assessment. The imaging configuration changes automatically according to the specific inspection scenario, optimizing the balance between detection precision and system complexity.
Solution Approach 2:
The patent implements parameter changes by varying imaging parameters (number of imaging points, angles, resolutions) based on the inspected region's risk level. High-risk areas receive imaging with multiple angles and higher resolution, while low-risk areas use fewer imaging points and lower resolution, thereby reducing overall system complexity while maintaining necessary precision.
3Measurement precision
If uniform imaging is performed across the entire range, then measurement precision is maintained, but loss of time increases due to unnecessary imaging in low-risk areas
Solution Approach 1:
The patent applies segmentation by dividing the workpiece inspection area into high-risk and low-risk regions based on historical flaw data and statistical analysis. Different imaging strategies are applied to each segment: comprehensive multi-angle imaging for high-risk segments and minimal imaging for low-risk segments. This segmentation eliminates unnecessary imaging time in low-risk areas while maintaining precision in critical regions.
Solution Approach 2:
The patent implements preliminary action by pre-processing historical flaw data to identify high-risk areas before actual inspection begins. The system uses past inspection results and flaw occurrence patterns to create a risk map that guides the inspection strategy, allowing the system to prioritize imaging resources on areas most likely to contain flaws.
Data Source
AI summary
A machine learning device that acquires state information from a robot control inspection system. The system has a robot hand to hold a workpiece or camera. The state information includes a flaw detection position of the workpiece, a movement route of the robot hand, an imaging point of the workpiece, and the number of imaging by the camera. A reward calculator calculates a reward value in reinforcement learning based on flaw detection information including the flaw detection position. A value function updater updates an action value function by performing the reinforcement learning based on the reward value, the state information, and the action.