Inference Calculation Processing Device for Robot Picking Efficiency
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Inference calculation processing for picking positions in a production line is inefficient due to unnecessary calculations on high-resolution distance images and three-dimensional point cloud data, leading to prolonged processing times and decreased production efficiency.
Innovation Solution
An inference calculation processing device and method that divides inference data into sub-data through batch processing, optimizing the calculation sequence based on evaluation scores to prioritize and perform calculations only on high-value data, thereby reducing unnecessary processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If inference calculation processing is performed on all clipped images (picking position candidates) with high-resolution distance images and three-dimensional point cloud data, then comprehensive evaluation of all candidates is achieved, but processing time is significantly lengthened and production efficiency decreases
Solution Approach 1:
The patent segments the inference processing into two distinct stages: a first inference calculation that performs comprehensive evaluation on all clipped images using high-resolution data to ensure accuracy, and a second inference calculation that performs rapid evaluation on selected candidates using low-resolution data to reduce processing time. This segmentation allows the system to maintain measurement precision for critical evaluations while minimizing time loss through efficient processing of subsequent evaluations.
Solution Approach 2:
The patent dynamically adjusts the resolution of distance images and three-dimensional point cloud data based on the processing stage. High-resolution data is used in the first inference calculation when accuracy is paramount, while low-resolution data is used in the second inference calculation when speed is prioritized. This dynamic adaptation allows the system to optimize the balance between evaluation accuracy and processing time according to specific operational requirements.
2Measurement precision
If high-resolution distance images and three-dimensional point cloud data are used for inference calculation, then evaluation accuracy is improved, but data size increases leading to longer processing times
Solution Approach 1:
The patent applies local quality by using high-resolution distance images and three-dimensional point cloud data only for the first inference calculation where accuracy is critical, while using low-resolution data for the second inference calculation where speed is more important. This localized application of data quality ensures that computational resources are concentrated where they provide the most value, maintaining evaluation accuracy for initial candidate selection while preserving processing throughput for subsequent evaluations.
Solution Approach 2:
The patent dynamically switches between high-resolution and low-resolution data based on the inference stage. The system transitions from high-resolution data in the first inference calculation to low-resolution data in the second inference calculation, allowing it to adapt to different performance requirements at different stages of the processing pipeline, thereby optimizing both accuracy and productivity.
3Reliability
If all picking position candidates are evaluated through inference calculation, then optimal picking position selection is ensured, but robot standby time increases reducing production efficiency
Solution Approach 1:
The patent segments the candidate evaluation process into two phases: a first inference calculation that reliably evaluates all clipped images to identify promising candidates, and a second inference calculation that quickly evaluates selected candidates to determine the final optimal picking position. This segmentation ensures that reliability is maintained in the initial screening while minimizing robot standby time through rapid final evaluation.
Solution Approach 2:
The patent performs preliminary action by conducting the first inference calculation on all clipped images before the robot needs to execute the picking operation. This preliminary evaluation identifies and prioritizes promising candidates, allowing the second inference calculation to focus only on selected candidates. This preliminary screening ensures reliable candidate identification while reducing the time the robot must wait for final evaluation results.
Data Source
AI summary
The present invention executes an inference calculation process in a short time without a robot having to wait for prolonged periods of time. This inference calculation processing device inputs inferencing data into a trained model and executes a process of inference calculation on the inferencing data, wherein the inference calculation processing device comprises: an acquisition unit for acquiring the inferencing data and the trained model; a preprocessing unit for batching the acquired inferencing data and dividing the batches into a plurality of inferencing sub-data; and an execution unit for optimizing the order of the process of inference-calculating the plurality of inferencing sub-data and executing the process of inference-calculating the inferencing data in the optimized order of the inference calculation process on the basis of each of at least some of the plurality of inferencing sub-data and the trained model.