Autonomous Workpiece Transfer With ML-Guided Pickability Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automated workpiece transfer systems struggle to efficiently identify and process diverse workpiece types, leading to reduced productivity and increased costs due to the need for customized and reconfigurable feeders.
Innovation Solution
The system employs an autonomous robot equipped with a transfer component, actuator components, and an end-of-arm-tooling component, in conjunction with a processor that applies a machine-learning model to identify workpiece types and adapt operating commands to increase the number of pickable workpieces transferred to a picking area.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If traditional feeders are used to transfer workpieces, then workpiece transfer can be achieved, but the system requires customized and reconfigurable feeders for diverse workpiece types, increasing device complexity and costs
Solution Approach 1:
The patent implements a universal autonomous robot system that can handle multiple workpiece types through machine learning identification and adaptive command generation. Instead of requiring customized feeders for each workpiece type, the system uses a single robot platform that automatically adapts to different workpiece types by identifying them via imaging and selecting appropriate reference operating commands, thereby achieving multi-functionality and reducing device complexity
Solution Approach 2:
The system enables the robot to autonomously identify workpiece types using machine learning models applied to images, automatically select appropriate reference operating commands, and self-adjust its transfer operations without human intervention. This self-service capability eliminates the need for manual reconfiguration of feeders when changing workpiece types, reducing both device complexity and operational overhead
2Productivity
If machine learning models are applied to identify workpiece types, then productivity is improved through automated identification, but system complexity increases due to the need for imaging and processing capabilities
Solution Approach 1:
The patent replaces traditional mechanical workpiece identification methods (such as physical sensors or manual inspection) with machine learning-based image recognition. The system captures images of workpieces and uses trained machine learning models to automatically identify workpiece types, substituting mechanical/electrical sensing systems with intelligent software-based identification that improves productivity while managing system complexity through software rather than hardware additions
3Productivity
If the number of pickable workpieces is increased through actuator components, then transfer efficiency is improved, but the system requires sophisticated control and adjustment mechanisms
Solution Approach 1:
The patent implements dynamic adjustment of actuator components based on real-time workpiece identification results. The system selects reference operating commands that specify actuator parameters (such as vibration amplitude, duration, or pattern) tailored to each identified workpiece type, enabling the actuators to dynamically adapt their behavior to maximize the number of pickable workpieces while managing complexity through software-controlled parameter adjustment rather than fixed mechanical configurations
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Automated workpiece transfer systems and methods of implementing thereof are disclosed. The system includes an autonomous robot and a processor in communication with the robot. The autonomous robot includes a transfer component operable to transfer the workpieces to a picking area, one or more actuator components operable to increase a number of pickable workpieces transferred to the picking area according to a set of actuator operating commands; and an end-of-arm-tooling component operable to retrieve pickable workpieces from the picking area and to transfer the pickable workpieces to a receiving area. The processor is operable to apply a workpiece identification machine-learning model to an image to identify a reference workpiece type associated with the workpieces shown within the image; identify a set of reference operating commands associated with the reference workpiece type; and define the set of actuator operating commands based on the image and the set of reference operating commands.