Autonomous Driving Model Training Using Driver and Sensor Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving technologies face limitations due to the limited recognition range and low reliability of sensors, as well as the complexity of real-road driving environments, making it difficult to define autonomous driving rules for all surrounding conditions.
Innovation Solution
An apparatus and method that collect driver information, such as gaze direction, heart rate, and driving patterns, along with sensor information like image and lidar data, to train an autonomous driving model using a deep neural network, enabling the model to learn about driving areas and dangerous situations, and integrate navigation information to train for specific routes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If autonomous driving technology depends only on sensors installed in a vehicle, then the system structure remains simple, but the recognition range is limited and reliability is low
Solution Approach 1:
The patent combines multiple information sources including sensor data from the vehicle, map data from navigation systems, and driver information from monitoring devices to create a comprehensive autonomous driving system. This merging of multiple data sources expands the recognition range and improves reliability while distributing system complexity across different functional modules.
2Adaptability or versatility
If autonomous driving rules are defined for all surrounding conditions, then the adaptability improves, but the complexity of defining and processing rules increases significantly
Solution Approach 1:
The patent segments the autonomous driving system into multiple independent modules: sensor information processing, map data processing, driver information processing, and integration layers. Each module handles specific aspects of driving conditions independently, allowing the system to adapt to various situations without requiring a single complex rule set to cover all possible conditions.
Solution Approach 2:
The patent introduces a new dimension by incorporating driver information (gaze direction, heart rate, driving patterns) alongside traditional sensor and map data. This additional dimension allows the system to adapt to different driving situations by considering driver state and behavior, reducing the need for exhaustive rule definitions for every possible condition.
3Measurement precision
If driver information and sensor information are collected and processed, then the accuracy and reliability of autonomous driving improves, but the amount of data processing and computational requirements increase
Solution Approach 1:
The patent performs preliminary processing of sensor information, map data, and driver information separately before integration. Each data type is pre-processed and filtered independently to extract relevant features, reducing the computational burden during real-time decision-making. This preliminary action allows the system to maintain high accuracy while managing computational requirements through distributed preprocessing.
Data Source
AI summary
Disclosed is an apparatus and method to train an autonomous driving model. The apparatus includes a driver information collection processor configured to collect driver information while a vehicle is being driven. The apparatus also includes a sensor information collection processor configured to collect sensor information from a sensor installed in the vehicle while the vehicle is being driven, and a model training processor configured to train the autonomous driving model based on the driver information and the sensor information.