Real-Time Predictive Model Control for Work Machine Yield Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for controlling work machines, such as combine harvesters, rely on a priori data for generating predictive models, which do not account for actual, ground truth data, leading to inaccurate yield predictions and ineffective real-time control during operations.
Innovation Solution
A system that generates a predictive model using both a priori geo-referenced vegetative index data and in situ field data collected during operations, dynamically evaluating and updating the model for improved accuracy and control, allowing for real-time adjustments and switching to alternative models if necessary.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a priori data is used to generate predictive models, then the system can provide initial control guidance, but the model accuracy deteriorates because it does not account for actual ground truth data
Solution Approach 1:
The system implements feedback by continuously comparing predicted values from the model with actual sensor measurements (ground truth data) during machine operation. This feedback loop enables the system to identify discrepancies and trigger model retraining, thereby improving model accuracy over time while incorporating real-time operational data.
Solution Approach 2:
The system performs self-service through automated model retraining and validation. When ground truth data becomes available during operation, the system automatically retrains the predictive model using this new information, validates the updated model, and deploys it without external intervention, thereby continuously improving itself based on actual performance data.
2Measurement precision
If the predictive model is updated in real-time during operations, then the model accuracy improves, but the system complexity increases due to dynamic model generation and validation requirements
Solution Approach 1:
The system embraces dynamics by allowing the predictive model to evolve from a static a priori model to a dynamic model that adapts during operation. The model structure and parameters are adjusted in real-time based on incoming ground truth data, enabling the system to respond to changing field conditions while maintaining improved prediction accuracy.
Solution Approach 2:
The system performs preliminary actions by pre-establishing the model validation framework and retraining mechanisms before operation begins. Validation rules, quality thresholds, and retraining protocols are configured in advance, allowing the system to efficiently process and validate new data during operation without adding operational complexity.
3Reliability
If the system uses actual field data collected during operations, then the predictive model becomes more accurate, but the time required for data collection and processing increases
Solution Approach 1:
The system maintains continuity of useful action by collecting and processing data continuously during machine operation rather than in discrete batches. Sensors continuously capture ground truth data, which is immediately fed into the model retraining process, ensuring that the model evolves continuously without interrupting machine operations or losing valuable real-time information.
Solution Approach 2:
The system applies partial action by selectively using only the portion of collected data that is most relevant and reliable for model retraining. Rather than processing all available data equally, the system identifies and utilizes high-quality ground truth measurements that provide the greatest improvement to model accuracy, thereby reducing processing time while maintaining reliability.
Data Source
AI summary
A priori geo-referenced data is obtained for a worksite, along with field data that is collected by a sensor on a work machine that is performing an operation at the worksite. A predictive model is generated, while the machine is performing the operation, based on the geo-referenced data and the field data. A model quality metric is generated for the predictive model and is used to determine whether the predictive model is a qualified predicative model. If so, a control system controls a subsystem of the work machine, using the qualified predictive model, and a position of the work machine, to perform the operation.