Robotic Arm Action Prediction for Separating and Grasping Stacked Objects

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Robotic arms struggle to accurately separate tightly stacked objects and successfully grasp them due to challenges in predicting effective actions.

Innovation Solution

A method involving the acquisition of depth and color images to predict probabilities of separating and grasping objects, using machine learning networks to determine optimal actions, and controlling the robotic arm accordingly.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If traditional robotic control methods are used, then the robotic arm can perform basic grasping actions, but it fails to accurately separate tightly stacked objects and successfully grasp them

Engineering Contradiction:

Improvegrasping success rateVSAvoidcontrol system complexity

Core Design Contradiction:

ReliabilityVSDevice complexity

Solution Approach 1:

The system performs preliminary actions by predicting the outcomes of multiple candidate actions before execution. It evaluates potential grasping actions in advance using deep learning models that analyze depth images and color images, determining which actions are most likely to succeed before the robotic arm executes them, thereby improving grasping success rate without increasing real-time control complexity

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system incorporates feedback mechanisms where the predicted values from candidate actions are used to determine the target action. The deep learning models continuously learn from the environment and adjust their predictions based on observed outcomes, creating a feedback loop that improves the robotic arm's grasping accuracy and reliability over time

Inventive Principle:

Principle #23Feedback

2Measurement precision

If multiple candidate actions are evaluated using deep learning models, then the accuracy of determining target actions is improved, but the computational time and processing requirements increase

Engineering Contradiction:

Improveaction prediction accuracyVSAvoidprocessing time

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system evaluates multiple candidate actions (excessive action) but only executes the most promising one (partial action). By predicting and filtering candidate actions in advance using deep learning models, the system processes more possibilities than it needs to execute, improving accuracy while managing processing time through selective execution of only the top predicted actions

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS12552043B2Method for controlling robotic arm, electronic device, and computer-readable storage medium

Publication Date: 2026.02.17 JD DIGITS HAIYI INFORMATION TECHNOLOGY CO LTD

AI summary

A method for controlling a robotic arm includes: acquiring a first depth image and a first color image of a target object; predicting first predicted values and second predicted values of respective candidate actions of the robotic arm based on the first depth image and the first color image, in which the first predicted value represents a probability of the robotic arm separating the target object from an adjacent object by executing the corresponding candidate action; and the second predicted value represents a probability of the robotic arm successfully grabbing the target object by executing the corresponding candidate action; determining a target action based on the first predicted values and the second predicted values of the candidate actions; and controlling the robotic arm to execute the target action.