Robot Grasp Planning Using Image-Based Evaluation Values
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current robot systems for object manipulation, such as picking automation in warehouses, face challenges in accurately determining the grasping position and posture of objects using sensor data, particularly in complex environments, leading to inefficiencies in object handling tasks.
Innovation Solution
An object manipulation apparatus equipped with a memory and hardware processor that calculates evaluation values for grasping approaches based on image data, generates information for optimal grasping strategies, and controls the actuation of grasping tools using machine learning algorithms, such as convolutional neural networks and reinforcement learning, to improve the efficiency and accuracy of object handling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional sensor data processing methods are used to determine grasping position and posture, then the system structure remains simple, but the accuracy of grasping determination deteriorates in complex environments
Solution Approach 1:
The patent replaces conventional sensor data processing methods with a deep learning-based image processing system. The object manipulation apparatus uses an image generator to create synthetic training images and a neural network to automatically determine grasping positions and postures, substituting traditional mechanical sensor processing with intelligent algorithm-based processing that achieves higher accuracy in complex environments.
2Measurement precision
If deep learning methods are implemented to improve grasping accuracy, then measurement precision improves, but loss of time increases due to complex processing requirements
Solution Approach 1:
The patent implements preliminary action by generating training images and training the neural network in advance before actual object manipulation tasks. The image generator creates synthetic training data beforehand, and the neural network is pre-trained to rapidly infer grasping positions and postures during real operations, significantly reducing processing time during actual use while maintaining high accuracy.
3Productivity
If conventional robot systems are used for object handling, then device complexity remains low, but productivity decreases due to inefficiencies in handling multiple objects
Solution Approach 1:
The patent implements universality by designing a neural network that can handle multiple types of objects with various shapes, sizes, and orientations using a single unified system. The object manipulation apparatus uses the same deep learning-based grasping determination system for diverse objects, eliminating the need for multiple specialized systems and improving overall productivity through multi-functional capability.
Data Source
AI summary
An object manipulation apparatus according to an embodiment of the present disclosure includes a memory and a hardware processor coupled to the memory. The hardware processor is configured to: calculate, based on an image in which one or more objects to be grasped are contained, an evaluation value of a first behavior manner of grasping the one or more objects; generate information representing a second behavior manner based on the image and a plurality of evaluation values of the first behavior manner; and control actuation of grasping the object to be grasped in accordance with the information being generated.