Picking Robot Self-Learning for Diverse Targets
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In picking systems, automating operations for large varieties of targets is costly and labor-intensive, as traditional teaching methods become impractical and inefficient.
Innovation Solution
A picking system comprising a picking robot, an operation unit for remote operation, a learning unit that learns the robot's movement during remote operation, and an assisting unit that automates or guides the operation based on the learning results, reducing the need for manual teaching and increasing efficiency.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If traditional teaching methods are used for automating picking operations, then automation capability is improved, but cost and labor increase significantly when dealing with large varieties of targets
Solution Approach 1:
The picking robot performs self-learning by autonomously observing and recording the operator's manual picking operations. The learning unit automatically captures movement data, gripper force data, and operation sequences without requiring external teaching intervention, enabling the system to teach itself how to perform picking tasks for new target types.
Solution Approach 2:
The system creates a digital copy of the operator's picking behavior through the learning unit, which records and stores operation data including movement trajectories, gripper positioning, and force application. This copied data is then replayed and executed by the picking robot to replicate human picking actions automatically.
2Manufacturing precision
If manual teaching is performed for each target type, then picking accuracy is improved, but time and productivity deteriorate due to extensive teaching requirements
Solution Approach 1:
The learning unit continuously records and stores picking operation data during normal manual operations, preparing the knowledge base in advance. When a new target type needs to be picked, the system can quickly retrieve and apply previously learned movement patterns and gripper parameters, eliminating the need for time-consuming teaching procedures for each new target type.
Solution Approach 2:
The system incorporates feedback mechanisms where the learning unit continuously monitors and records actual picking operations, including successful and unsuccessful attempts. This feedback data is used to refine and improve the stored operation patterns, enabling the robot to learn from experience and improve picking accuracy over time without additional teaching intervention.
3Ease of operation
If full automation is implemented for diverse targets, then labor reduction is achieved, but system complexity and cost increase
Solution Approach 1:
The system implements partial automation where the learning unit and assisting unit handle the complex learning and guidance functions, while the picking robot executes the learned operations. This partial automation approach reduces labor for the most complex aspects (learning and adaptation) while maintaining simplicity in the execution layer, avoiding the need for fully complex autonomous systems.
Solution Approach 2:
The learning unit and assisting unit serve as intermediaries between the operator and the picking robot. Instead of direct full automation, these intermediary components capture operator knowledge, process it into actionable instructions, and guide the robot's operations, thereby reducing labor while managing system complexity through a layered architecture.
Data Source
AI summary
A picking system including: a picking robot for gripping a target; an operation unit for an operator to perform a remote operation of the picking robot; a learning unit that learns a movement of the picking robot when the target is gripped by the remote operation; and an assisting unit that assists the remote operation based on a learning result of the learning unit.