Neural Network Friction Coefficient Retraining
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Training deep neural networks for autonomous vehicles to determine coefficients of friction requires extensive and costly data acquisition and processing, as they struggle with unseen scenarios, leading to inefficiencies in retraining and accuracy.
Innovation Solution
A system that determines when a deep neural network encounters an unseen coefficient of friction scenario by comparing image data with a threshold, allowing for retraining using additional vehicle sensor data and enabling real-time or near-real-time updates to improve accuracy and adaptability.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep neural networks are trained with extensive data acquisition to handle unseen scenarios, then accuracy improves, but time consumption and cost increase significantly
Solution Approach 1:
The system performs preliminary actions by continuously monitoring and detecting unseen scenarios during normal operation, preparing retraining data in advance. When an unseen scenario is detected, the system has already collected relevant sensor data and images, so the actual retraining process can start immediately without waiting for extensive data acquisition, thus reducing time consumption while maintaining accuracy.
Solution Approach 2:
The system implements a feedback mechanism where the neural network's predictions are continuously evaluated against actual sensor data. When discrepancies indicate unseen scenarios, the system triggers retraining using the collected feedback data. This closed-loop feedback allows the system to improve accuracy through targeted retraining rather than extensive pre-training, reducing the time and cost burden.
2Adaptability or versatility
If deep neural networks are retrained frequently to adapt to new scenarios, then adaptability improves, but computational resources and processing time increase
Solution Approach 1:
Instead of performing full extensive data acquisition and complete retraining processes, the system applies partial action by using only the specific sensor data and images collected during the unseen scenario detection. This targeted approach provides sufficient adaptability improvement without the excessive computational resources required for complete retraining pipelines.
Solution Approach 2:
The system enables self-service by automatically detecting unseen scenarios, collecting relevant data, and triggering retraining without external intervention. This automated self-service mechanism improves adaptability continuously while minimizing the need for external computational resources and manual processing.
3Reliability
If extensive data acquisition is performed to cover all possible scenarios, then reliability improves, but device complexity and cost increase
Solution Approach 1:
The system extracts only the essential and relevant sensor data and image information needed for detecting unseen scenarios and performing targeted retraining. By taking out only the necessary data elements rather than acquiring extensive comprehensive datasets, the system achieves reliable adaptation to new scenarios while reducing device complexity and associated costs.
Data Source
AI summary
A system comprising a computer including a processor and a memory, the memory including instructions such that the processor is programmed to: determine whether a difference between a friction coefficient label and a determined friction coefficient corresponding to an image depicting a surface is greater than a label threshold; modify the determined friction coefficient to equal the friction coefficient label when the difference is greater than the label threshold; and retrain a neural network using the image and the friction coefficient label.