Adaptive Deep Learning Model for Autonomous Vehicles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles use a single deep learning model for all drivers, which limits their ability to adapt to individual driving tendencies and preferences, resulting in suboptimal performance.
Innovation Solution
A method and device for providing dynamic, adaptive deep learning models tailored to individual users by fine-tuning models using video data from autonomous vehicles, incorporating convolutional operations, backpropagation, and personalized policies, with a system for updating and validating labeled data to optimize autonomous driving.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a single deep learning model is used for all drivers, then device complexity is reduced and ease of manufacture is improved, but adaptability to individual driving tendencies deteriorates
Solution Approach 1:
The patent segments the single deep learning model into multiple user-specific models, where each model is customized for individual driving tendencies. The system divides the general model into user-specific instances through fine-tuning processes, allowing each driver to have a personalized model while maintaining the underlying shared architecture.
Solution Approach 2:
The patent applies parameter changes by fine-tuning model parameters for each user based on their driving behavior data. The system modifies specific parameters of the deep learning model to adapt to individual driving tendencies, while keeping the overall model structure unchanged, thus achieving personalization without complete model recreation.
2Reliability
If deep learning models are fine-tuned for each user, then personalized performance is improved, but loss of time for data processing and model updating increases
Solution Approach 1:
The patent implements preliminary action by pre-collecting and storing driving behavior data during normal operation. The system accumulates user-specific driving patterns in advance, preparing the data needed for future fine-tuning operations. This allows the model to be updated efficiently when conditions are appropriate, rather than processing everything in real-time.
Solution Approach 2:
The system employs periodic action by updating user-specific models at predetermined intervals or when sufficient new data is accumulated. Rather than continuous real-time fine-tuning, the system periodically processes and updates models, balancing performance improvement with computational efficiency and time management.
3Measurement precision
If video data is collected and processed for each user, then measurement precision of driving tendencies is improved, but quantity of data to be processed increases
Solution Approach 1:
The patent applies taking out by extracting only the essential features and patterns from video data that are relevant to driving tendencies. The system processes video data to extract meaningful behavioral patterns while discarding redundant information, thus achieving precise measurement without processing the entire raw video dataset.
Solution Approach 2:
The system implements partial action by focusing on specific aspects of driving behavior that are most indicative of user tendencies. Rather than analyzing all possible video data comprehensively, the system selectively processes key features and patterns that provide sufficient precision for personalization while reducing overall data processing requirements.
Data Source
AI summary
A method for providing a dynamic adaptive deep learning model other than a fixed deep learning model, to thereby support at least one specific autonomous vehicle to perform a proper autonomous driving according to surrounding circumstances is provided. And the method includes steps of: (a) a managing device which interworks with autonomous vehicles instructing a fine-tuning system to acquire a specific deep learning model to be updated; (b) the managing device inputting video data and its corresponding labeled data to the fine-tuning system as training data, to thereby update the specific deep learning model; and (c) the managing device instructing an automatic updating system to transmit the updated specific deep learning model to the specific autonomous vehicle, to thereby support the specific autonomous vehicle to perform the autonomous driving by using the updated specific deep learning model other than a legacy deep learning model.