Excavation Planning Using Soil-Aware ML Trajectory Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing excavation planning methods do not adequately consider the influence of soil quality, leading to inefficiencies in excavated soil amount and work time variations due to soil type variations.
Innovation Solution
An excavation plan creation device utilizing machine learning models that incorporate soil quality parameters to determine excavation trajectories and swing directions, including a soil quality estimation unit, terrain information acquisition, and planned value calculation to optimize excavation plans based on soil type.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If excavation planning is performed using fixed rules based on veteran operator experience, then the excavation plan can be created with simple methodology, but the plan cannot appropriately adapt to different soil qualities leading to work efficiency variations
Solution Approach 1:
The patent applies parameter changes by transitioning from fixed rule-based planning to machine learning models that dynamically adjust excavation trajectories and swing directions based on soil quality parameters. The system learns optimal parameters from training data and adapts to different soil conditions (sand, clay, etc.) by selecting appropriate pre-trained models or adjusting model parameters based on detected soil quality.
Solution Approach 2:
The patent replaces the mechanical system of fixed rule-based decision making with an intelligent system using machine learning models. Instead of relying on static expert knowledge rules, the system uses trained neural networks or other ML algorithms that can process soil quality data and generate optimized excavation plans adaptable to various soil types.
2Reliability
If soil quality parameters are partially considered in fixed-rule planning, then the planning methodology remains simple, but the consideration of soil quality influence remains insufficient
Solution Approach 1:
The patent replaces simple rule-based soil quality consideration with machine learning models that automatically learn and apply soil quality influences from training data. The ML models process soil quality parameters and generate optimized excavation trajectories without requiring complex manual rule formulation, achieving high reliability through data-driven learning.
Solution Approach 2:
The system applies self-service by using machine learning models that automatically adapt to different soil qualities without requiring manual reconfiguration of rules. The models are pre-trained on diverse soil conditions and can independently select or adjust their behavior based on detected soil quality, reducing the need for complex external control mechanisms.
3Productivity
If different excavation plans are created for different soil qualities, then work efficiency can be optimized, but the complexity of plan creation increases
Solution Approach 1:
The patent applies universality by developing a unified machine learning-based planning system that handles multiple soil qualities through a single framework. Instead of creating separate planning systems for each soil type, the system uses versatile ML models that can process various soil quality inputs and generate optimized plans across different conditions, reducing overall system complexity.
Solution Approach 2:
The system manages complexity by using parameter changes in the machine learning models themselves rather than changing the entire planning methodology for each soil type. The models adjust their output trajectories and parameters based on input soil quality data, allowing efficient adaptation to different soils through parameter adjustment rather than structural complexity.
Data Source
AI summary
An excavation plan creation device includes a storage unit configured to store a plurality of planning models each of which is a machine learning model that receives, as an input, terrain information and that outputs a planned value of an excavation trajectory and a swing direction. The planning models have performed machine learning with parameters related to soil quality which are different from each other. A soil quality estimation unit is configured to estimate soil quality. A terrain information acquisition unit is configured to acquire terrain information. A planned value calculation unit is configured to select the planning model on the basis of the soil quality estimated by the soil quality estimation unit, input the terrain information acquired by the terrain information acquisition unit to the selected planning model, and calculate the planned value as an output of the planning model.