Robot Grasping Sequence Optimization via Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods for a robot to grasp and store objects on a conveyor fail to optimize the grasp order based on object positions and orientations, leading to inefficient movement and potential missed container opportunities due to excessive rotation and varying cycle times, torque, and vibration, which are not effectively managed by existing systems.
Innovation Solution
A machine learning device that observes object positions and orientations, calculates state variables such as cycle time, torque, and vibration, and learns optimal grasp orders using a combination of state observation, determination data, and reinforcement learning to minimize cycle time and reduce mechanical stress.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If objects are grasped in the order they are located downstream on the conveyor, then the grasping process is simple to implement, but the hand rotates significantly and movement time varies widely
Solution Approach 1:
The patent applies dynamics by making the grasping order flexible and adaptive rather than fixed. The system dynamically determines the optimal grasping sequence based on real-time object positions, orientations, and container locations, allowing the robot hand to minimize rotation and movement time while adapting to varying conveyor configurations
Solution Approach 2:
The patent changes the parameter of grasping order from a fixed downstream sequence to a variable sequence optimized for each specific configuration. By calculating and adjusting the grasping sequence based on object positions, orientations, and container locations, the system reduces hand rotation and movement time while maintaining ease of implementation through automated optimization
2Length of moving object
If the conveyor is divided into smaller regions to reduce carry distance, then the movement distance is reduced, but the system complexity increases and effect is limited when conveyor is wide
Solution Approach 1:
The patent moves beyond simple spatial division of the conveyor into regions and instead optimizes across multiple dimensions simultaneously - considering object positions, orientations, container locations, and robot kinematics. This multi-dimensional optimization approach reduces carry distance more effectively than regional division while avoiding increased system complexity through integrated calculation
3Ease of operation
If the robot grasps objects without considering orientations, then the grasping process is simpler, but the robot may grasp objects with unfavorable postures requiring significant rotation
Solution Approach 1:
The patent applies preliminary action by calculating and determining the optimal grasping sequence and orientation adjustments in advance, before the robot executes the grasping operations. The system pre-processes object position and orientation data to determine the most efficient grasping sequence that minimizes rotation, maintaining simplicity while reducing rotation time through advance planning
4Reliability
If the robot stops the conveyor for containers to grasp objects, then all objects can be grasped, but production volume cannot be met and conveyance cannot be stopped
Solution Approach 1:
The patent maintains continuity of useful action by enabling the conveyor to continue moving while the robot grasps objects. The optimized grasping sequence allows the robot to efficiently pick multiple objects from the moving conveyor and place them in passing containers without stopping the conveyor, ensuring both complete object grasping and continuous production flow
Data Source
AI summary
A machine learning device according to the present invention learns an operation condition of a robot that stores a plurality of objects disposed on a carrier device in a container using a hand for grasping the objects. The machine learning device includes a state observation unit for observing the positions and postures of the objects and a state variable including at least one of cycle time to store the objects in the container and torque and vibration occurring when the robot grasps the objects during operation of the robot; a determination data obtaining unit for obtaining determination data for determining a margin of each of the cycle time, the torque, and the vibration against an allowance value; and a learning unit for learning the operation condition of the robot in accordance with a training data set constituted of a combination of the state variable and the determination data.