Active Ground Terrain Mapping for Uneven-Field Header Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems for predicting ground elevation in front of mobile machines, such as combine harvesters, using past elevation measurements are inaccurate, especially over uneven terrain, leading to significant errors in header height control and crop volume calculations.
Innovation Solution
An elevation map generator that estimates elevation values for points near the machine based on a plane derived from a pose detection system, with confidence values assigned based on distance, allowing for real-time aggregation and use in controlling controllable subsystems like header height.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a plane is projected forward from past elevation measurements to predict ground elevation, then the system can provide ground elevation data without additional sensors, but the prediction accuracy deteriorates significantly over uneven terrain and at distances further from the machine
Solution Approach 1:
The system performs preliminary mapping by collecting and processing elevation data as the machine moves through the field, building a point cloud and generating terrain models in advance. This preliminary action creates a foundation of known elevation data that can be used to predict ground elevation in areas not yet directly measured, reducing reliance on simple plane projections while maintaining system complexity at acceptable levels.
Solution Approach 2:
The system creates a digital copy of the terrain through point cloud representation and terrain models. This digital copy allows the system to store, process, and reference elevation data without requiring physical measurement devices at every location. The copied terrain information can be interpolated and projected forward with much higher accuracy than simple plane projections, solving the contradiction between system simplicity and measurement precision.
2Length of stationary object
If elevation data is collected from distant points ahead of the machine, then the prediction range is extended, but the accuracy of elevation estimates decreases due to greater uncertainty
Solution Approach 1:
The system applies different quality levels to different regions of the terrain model. Areas with dense point cloud data receive higher confidence weights, while distant or sparsely sampled areas receive lower confidence weights. This local quality approach allows the system to extend prediction range while maintaining high accuracy in well-sampled regions and appropriately reduced expectations in distant regions, resolving the contradiction between prediction range and accuracy.
Solution Approach 2:
The system continuously refines elevation estimates by incorporating new measurement data as the machine moves forward. Previously uncertain distant points become nearer and are re-evaluated with updated information. This feedback mechanism allows the system to extend prediction range while progressively improving accuracy over time through iterative refinement of the terrain model based on accumulating data.
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
An elevation map generator in a mobile agricultural machine generates an elevation map by estimating an elevation value for points in front of a work machine based on a plane derived from a measured elevation point measured by a pose detection system affixed to the work machine. Each elevation value has a corresponding confidence value that varies inversely with a distance of the point from the pose detection system. As the machine moves, additional elevation values are aggregated for each point, based on the confidence values. The machine is controlled based on the aggregated evaluation values.