Field Route Planning Using Deep Learning for Mobile Machines
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing route planning systems for mobile machines lack the sophistication to consider complex interactions between terrain, ground conditions, soil type, weather conditions, and machinery capabilities, resulting in subpar paths and increased fuel consumption, leading to higher operational costs and reduced productivity.
Innovation Solution
The use of machine learning, specifically deep learning models based on Recurrent Neural Networks (RNNs) or Long Short-Term Memory (LSTM) networks, to generate routing plans for mobile machines, taking into account various factors such as terrain, ground conditions, soil type, weather conditions, and machinery capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If manual planning or simple heuristic algorithms are used for route planning, then the system complexity is low and ease of operation is maintained, but the route optimization quality deteriorates and fuel consumption increases
Solution Approach 1:
The patent replaces manual planning and simple heuristic algorithms with a deep learning-based computational system. The system uses neural networks to process complex interactions between terrain, ground conditions, soil type, weather, and machinery capabilities, automatically generating optimized routes without requiring manual intervention or simple rule-based computations.
Solution Approach 2:
The patent transforms the route planning approach by changing from discrete heuristic rules to continuous deep learning model predictions. The system processes multiple environmental and machinery parameters simultaneously, generating optimized routes that consider complex interactions between these parameters rather than applying simple individual rules.
2Productivity
If complex interactions between terrain, ground conditions, soil type, weather, and machinery capabilities are considered, then route optimization quality improves and fuel consumption decreases, but the computational complexity and data processing requirements increase
Solution Approach 1:
The patent applies preliminary action by training deep learning models offline with historical data containing complex interactions between terrain, ground conditions, soil type, weather, and machinery capabilities. Once trained, the model can quickly generate optimized routes during operation without requiring complex real-time computations, thus achieving fuel efficiency without excessive operational computational complexity.
Solution Approach 2:
The patent uses copying by creating a digital twin or virtual representation of the field environment and machinery characteristics. The deep learning model learns from copied historical routing data and environmental conditions, generating optimized routes based on patterns learned from previous operations rather than requiring complex real-time analysis of all parameters.
3Measurement precision
If deep learning models are trained with historical routing data, then the route planning accuracy improves and operational costs decrease, but the initial data processing time and computational resources increase
Solution Approach 1:
The patent performs the computationally intensive data processing and model training as a preliminary action before actual route planning operations. Historical routing data is processed and models are trained in advance, allowing the system to achieve high route planning accuracy during operation without incurring time losses during actual field work. The one-time training investment enables rapid subsequent route generation.
Data Source
Figure 1
Figure 2
Figure 3~5
AI summary
Embodiments include technologies that use machine learning to plan a route. Some embodiments include a method that includes using machine learning to generate or update a route plan. In some examples, the method includes receiving, by a computing system (102, 200), initial routing information (104), the initial routing information defining routes followed by or to be followed by one or more mobile machines in one or more fields (step 302). In such examples, the method also includes training, by the computing system(102, 200), a deep learning model (106) using the initial routing information (104) (step 302). Also, in such examples, the method includes using, by the computing system(102, 200), the trained model (107) to generate new routing information (108) for a given field (step 306).