Tractor Predictive Control Using Online Learning for Field Inertia

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Agricultural working machines, such as tractors, face challenges in achieving optimal working results due to system inertia, which limits timely adjustments of settings during harvesting processes, especially when dealing with diverse applications and unknown environmental conditions.

Innovation Solution

The implementation of an online learning algorithm and model within the agricultural working machine's control device allows for real-time processing and adaptation of operating parameters, enabling predictive control of work processes based on data from adjacent field areas, optimizing machine parameters and strategies for improved performance and efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Manufacturing precision

If real-time detection and control of operating parameters is implemented, then working result quality is improved, but system response time is insufficient due to inherent system inertia

Engineering Contradiction:
Improveworking result qualityVSAvoidsystem response time
Core Design Contradiction:
Manufacturing precisionVSLoss of time

Solution Approach 1:

The system performs preliminary detection of operating parameters in advance before the actual working point is reached. By detecting parameters ahead of time and using predictive algorithms to forecast future states, the control system can prepare adjustment commands in advance, thereby compensating for the inherent system inertia and ensuring timely response when the working point is reached.

Inventive Principle:
Principle #10Preliminary action

2Device complexity

If limited sensor data from adjacent lanes is used, then system complexity is reduced, but working result optimization is insufficient for diverse applications

Engineering Contradiction:
Improvesystem complexityVSAvoidworking result optimization
Core Design Contradiction:
Device complexityVSManufacturing precision

Solution Approach 1:

A predictive algorithm acts as an intermediary between the limited sensor data and the control decisions. The algorithm processes and interprets the available data from adjacent lanes, filling in gaps and predicting parameters that cannot be directly measured, thereby enabling optimized control for diverse applications without requiring complex sensor systems.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Adaptability or versatility

If predictive control algorithms are implemented, then adaptability to new circumstances is improved, but computational requirements and processing time increase

Engineering Contradiction:
Improveadaptability to new circumstancesVSAvoidcomputational processing time
Core Design Contradiction:
Adaptability or versatilityVSLoss of time

Solution Approach 1:

The system implements predictive control at strategically selected working points rather than continuously across the entire field. By focusing computational resources on critical decision points where predictive control provides the most value, the system achieves good adaptability while keeping processing time and computational load within acceptable limits.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3626038B1Agricultural machine
Publication Date: 2022.03.09 CLAAS TRACTOR
  • EP3626038B1 patent drawingFigure 1
  • EP3626038B1 patent drawingFigure 2

AI summary

The invention relates to an agricultural machine, in particular a tractor, with at least one working unit (2) for performing or supporting agricultural work, wherein the agricultural machine (1) has a sensor (4) for acquiring operating parameters and a control unit (5) for controlling work processes of the agricultural machine (1). It is proposed that the control unit (5) be configured to perform predictive control of at least one work process of the agricultural machine (1) during the execution of an agricultural work task, based on information generated by at least one online learning algorithm and/or at least one online learning model based on the acquired operating parameters.