Unm Driving Model Training With Semi-Supervised Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current unmanned driving behavior decision-making models trained through supervised learning require extensive labeled sample data, consuming significant human resources and struggling with low accuracy due to data expansion difficulties.
Innovation Solution
The method employs semi-supervised learning with manifold dimension reduction to extract feature vectors from sample and target data, enabling efficient training of decision-making models without the need for massive labeled data, using an acquisition module, extraction module, and training module to generate and update feature vectors and decision-making models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If supervised learning is used to train the unmanned driving decision-making model, then the model training can be performed with a clear learning framework, but a mass of sample data needs to be collected and labeled which consumes huge human resources and results in low model training efficiency
Solution Approach 1:
The system performs self-labeling through the semi-supervised learning framework where the model learns from both labeled and unlabeled data, automatically utilizing the structure and patterns in unlabeled data to improve training without requiring extensive manual annotation, thus reducing human resource consumption while maintaining training effectiveness
Solution Approach 2:
The patent changes the training parameter from purely labeled data to a combination of labeled and unlabeled data through semi-supervised learning, fundamentally altering how the model learns and improving training efficiency by leveraging the abundance of unlabeled data available in unmanned driving scenarios
2Ease of manufacture
If supervised learning is used to train the model, then the training process can be straightforward, but the sample data is difficult to be expanded which results in low accuracy of the trained model in behavior decision-making
Solution Approach 1:
The semi-supervised learning framework serves multiple functions simultaneously: it maintains the simplicity of supervised learning for labeled data while adding the capability to utilize unlabeled data for model improvement, thus achieving both ease of implementation and enhanced decision-making accuracy through multi-functional training approach
Solution Approach 2:
The patent adds another dimension to the training data by incorporating unlabeled data alongside labeled data, expanding the data dimension from a single labeled dataset to a two-dimensional space of labeled and unlabeled data, thereby improving model accuracy without complicating the fundamental training process
3Measurement precision
If manifold dimension reduction is used to extract feature vectors, then the feature extraction can capture complex data structures, but the computational complexity increases
Solution Approach 1:
The patent extracts only the essential and discriminative features from the complex unmanned driving data using manifold dimension reduction, separating the critical feature information from the redundant data, thus achieving accurate feature representation while reducing unnecessary computational complexity
Data Source
AI summary
Various embodiments a method and apparatus for unmanned driving behavior decision-making and model training, and an electronic device. The method includes: acquiring sample data, wherein the sample data includes a sample image; extracting a sample feature vector corresponding to the sample data, wherein a feature vector of the sample image is extracted by manifold dimension reduction; and based on the sample feature vector, training by semi-supervised learning to obtain a target decision-making model, wherein the target decision-making model is used for decision-making classification.