Loading Preparation Timing Prediction via Primitive Action Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies struggle to predict the end timing of loading preparation from a backhoe to a dump truck, necessitating a technique for accurately determining the completion of a predetermined movement action.
Innovation Solution
A timing prediction method involving frequency analysis of movement data to extract a transition sequence of primitive actions, using Hidden Markov Models to segment and analyze the occurrence rate of these actions, and predict the end timing based on an action model.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If frequency analysis and Hidden Markov Models are used to predict end timing of loading preparation, then prediction accuracy is improved, but device complexity increases
Solution Approach 1:
The loading preparation action is segmented into multiple primitive actions (approach, position, lift) through frequency analysis and Hidden Markov Model segmentation. This allows the system to predict the end timing by analyzing the completion sequence of these segmented primitive actions, thereby improving prediction accuracy while managing complexity through structured decomposition
Solution Approach 2:
The patent introduces an action model as an intermediary that stores pre-learned transition sequences of primitive actions. This action model serves as a mediator between the complex frequency analysis/HMM processing and the final prediction output, enabling accurate timing prediction while simplifying the real-time processing complexity
2Reliability
If movement data from multiple positions is acquired and analyzed, then prediction reliability is improved, but loss of time increases
Solution Approach 1:
The system performs preliminary learning to build the action model storing transition sequences of primitive actions before actual prediction. This preliminary action allows the system to use pre-computed models during real-time prediction, improving reliability through comprehensive data analysis while reducing processing time by avoiding repeated complex computations
Solution Approach 2:
The patent dynamically adjusts the analysis process by using frequency analysis to identify relevant frequency bands and applying HMM segmentation adaptively. This dynamic approach allows the system to process multi-position movement data efficiently by focusing computational resources on the most informative segments, thereby improving reliability without excessive time loss
Data Source
AI summary
Movement data of a plurality of positions on an observation object performing a predetermined movement action is acquired, feature data is generated by performing frequency analysis on the movement data, a transition sequence of a primitive action is extracted by segmenting the feature data, and based on the transition sequence and an action model obtained by learning a transition sequence of a primitive action obtained by segmenting a movement state of a training object performing a movement action, an occurrence rate of the primitive action in the predetermined movement action is analyzed to extract a pattern of the primitive action, thereby predicting an end timing of the predetermined movement action.