Robotic Peg-in-Hole Insertion for Unknown Tilt Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional industrial robots lack adaptivity and flexibility to perform peg-in-hole tasks with components of unknown tilt, requiring intensive human labor for programming due to varying manufacturing conditions.
Innovation Solution
A robotic system employing force sensors, a control system, and machine learning techniques to apply forces, normalize time-series data, create a time-series prediction model, and calculate a matching ratio to determine optimal insertion strategies for components with unknown tilt.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional industrial robots are used for peg-in-hole tasks, then high repeatability is achieved, but adaptability and flexibility are lacking due to inability to handle components with unknown tilt
Solution Approach 1:
The robotic system employs force sensors to detect contact forces during insertion and uses machine learning models to process this feedback information. The system continuously monitors force magnitudes and directions, compares them against predicted patterns, and adjusts the insertion trajectory in real-time to accommodate components with unknown tilt, thereby maintaining both repeatability and adaptability
Solution Approach 2:
The machine learning model enables the robotic system to autonomously determine the optimal insertion trajectory without requiring pre-programming or manual intervention for each component variation. The system self-adjusts by processing force sensor data and selecting trajectories from the database based on real-time conditions, eliminating the need for intensive human programming while handling manufacturing variations
2Adaptability or versatility
If intensive human programming is applied to handle manufacturing variations, then adaptability is improved, but productivity and production cycle time are reduced
Solution Approach 1:
The system pre-establishes a database of insertion trajectories through offline machine learning training before actual production. During manufacturing, the system rapidly queries and selects appropriate pre-computed trajectories based on real-time force sensor feedback, avoiding the need for time-consuming online programming or adjustment while maintaining high adaptability to component variations
Solution Approach 2:
The patent replaces manual programming and mechanical adjustment processes with an automated machine learning-based decision system. The control system automatically processes force sensor data, determines optimal trajectories, and executes insertion operations without human intervention, significantly reducing production cycle time while maintaining adaptability to manufacturing variations
3Extent of automation
If force sensors and machine learning techniques are employed to handle unknown tilt, then adaptability and automation are improved, but device complexity increases
Solution Approach 1:
The machine learning model serves multiple functions: it predicts contact forces, determines optimal insertion trajectories, and adapts to different component geometries. The force sensors not only detect contact forces but also provide feedback for trajectory adjustment. This multi-functionality reduces the need for separate specialized systems, managing complexity while enhancing automation
Solution Approach 2:
The system manages complexity by focusing on key parameters (force magnitude, force direction, trajectory selection) rather than controlling all possible degrees of freedom. The machine learning model processes force sensor data and translates it into trajectory adjustments, simplifying the control problem by parameter transformation and pattern recognition rather than complex multi-variable control
Data Source
AI summary
A computer-implemented method executed by a robotic system for performing a positional search process in an assembly task is presented. The method includes applying forces to a first component to be inserted into a second component, detecting the forces applied to the first component by employing a plurality of force sensors attached to a robotic arm of the robotic system, extracting training samples corresponding to the forces applied to the first component, normalizing time-series data for each of the training samples by applying a variable transformation about a right tilt direction, creating a time-series prediction model of transformed training data, applying the variable transformation with different directions for a test sample, and calculating a matching ratio between the created time-series prediction model and the transformed test sample.