Robot Gripper Tire Picking with Vision-Based 3D Reconstruction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current tire picking technologies struggle with unpredictable tire arrangements, especially in heterogeneous bulk loads, where the order and dimensions of tires are unknown, and require specialized hardware and precise CAD data, limiting their adaptability and efficiency in handling diverse tire configurations.
Innovation Solution
A computer-implemented process using a combination of vision techniques and unsupervised learning to reconstruct 3D scenes from scattered point clouds, allowing a robot gripper to identify and pick tires based on minimal surface information, regardless of their arrangement, and adapt to various tire sizes and shapes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional laser scanning and CAD model matching methods are used to pick tires, then the system can achieve precise tire identification in controlled environments, but the system fails to adapt to unexpected or heterogeneous tire arrangements and requires highly specialized hardware facilities
Solution Approach 1:
The patent replaces traditional mechanical laser scanning systems with a vision-based approach using cameras and machine learning algorithms. The system uses image capture and neural network processing to identify tires, eliminating the need for complex mechanical scanning hardware and CAD model matching infrastructure.
Solution Approach 2:
The system employs unsupervised learning algorithms that automatically adapt to different tire arrangements without requiring pre-programmed knowledge or manual configuration. The neural network self-adjusts to identify tires in heterogeneous bulk loads, making the system self-sufficient across varying conditions.
2Reliability
If the system requires complete visibility and homogeneous tire arrangements for CAD matching, then model accuracy is maintained, but the system cannot handle heterogeneous bulk loads with unknown tire orders and dimensions
Solution Approach 1:
The patent changes the fundamental parameters of the identification system by transitioning from deterministic CAD model matching to probabilistic machine learning classification. This allows the system to handle variable tire parameters (sizes, shapes, arrangements) by learning patterns from training data rather than requiring exact model matches.
Solution Approach 2:
The system dynamically adapts to different tire configurations through trained neural networks that can process and classify various tire types, arrangements, and conditions. The identification approach evolves from static model matching to dynamic pattern recognition.
3Measurement precision
If specialized hardware facilities and precise CAD data are required for tire picking, then identification accuracy is improved, but device complexity and hardware requirements increase significantly
Solution Approach 1:
The patent replaces complex mechanical laser scanning facilities and CAD data infrastructure with simpler vision-based hardware (cameras) and software-based machine learning processing. This substitution maintains detection accuracy while dramatically reducing hardware complexity and infrastructure requirements.
Solution Approach 2:
Instead of requiring physical CAD models and precise geometric representations, the system creates digital representations through image capture and machine learning feature extraction. The neural network learns to identify tires from visual patterns rather than requiring exact geometric copies.
4Measurement precision
If the system uses fixed reference learning in controlled environments, then initial tire identification is accurate, but the system cannot manage unexpected situations or variations in tire arrangements
Solution Approach 1:
The system uses unsupervised learning algorithms that automatically adapt to new tire arrangements without requiring reprogramming or manual intervention. The neural network self-adjusts its parameters based on training data, enabling it to handle unexpected configurations while maintaining identification accuracy.
Solution Approach 2:
The identification system transitions from static reference learning to dynamic adaptive learning through trained neural networks. The system can process and adapt to varying tire conditions, arrangements, and types while maintaining consistent performance through learned patterns rather than fixed rules.
Data Source
Figure 1~2
Figure 3~4
Figure 5~6
AI summary
The invention is directed to a process for directing the movement of a robot (102) with a gripper (108) that picks a target tire (P*) from an arrangement of tires, characterized in that the process includes the following steps: - a step of providing a system (100) for picking one or more tires stored in the arrangement of tires, the robot forming part of this system; - a step of determining one or more parameters of the target tire in the arrangement of tires; - a step of determining the diameter of the target tire; - a step of determining the tire radius (RP) and the rim radius (RJ) of the target tire; - an approaching step whereby the robot approaches the target tire identified for picking; - a picking step whereby the target tire is picked by the gripper; and - a step of extracting the target tire from the arrangement of tires in order to place it in a target location.