Adaptive Autonomous Driving Model for New Scene Handling
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving vehicles face limitations in adapting to new scenes, leading to a lack of full automation across various conditions, requiring driver intervention when encountering unfamiliar environments.
Innovation Solution
An adaptive processing method that involves obtaining scene data, creating a test set, updating pre-training model parameters through gradient iteration to generate a scene model capable of outputting autonomous driving strategies for new scenes, allowing vehicles to adapt and continue driving autonomously.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If autonomous driving vehicles use pre-training models for prescribed scenes, then autonomous driving can be achieved under certain conditions, but the vehicle cannot adapt to new scenes and requires driver takeover
Solution Approach 1:
The system dynamically adapts the pre-training model to new scenes by updating model parameters based on scene data from unfamiliar environments. This allows the autonomous driving system to transition from static pre-trained models to dynamic adaptive models that can handle both prescribed and new scenes, resolving the contradiction between reliability in known scenes and adaptability to new scenes.
Solution Approach 2:
The system changes the parameters of the pre-training model when encountering new scenes. By detecting scene type and updating model parameters based on new scene data, the system maintains reliable autonomous driving in known scenes while adapting to new scenarios, thereby resolving the contradiction between reliability and adaptability.
2Reliability
If the vehicle converts from autonomous driving to manual driving when encountering new scenes, then driver safety is ensured, but full automation cannot be achieved and driver intervention is required
Solution Approach 1:
The system dynamically adjusts the level of automation based on scene recognition. When new scenes are detected, the system adapts the pre-training model to maintain autonomous driving capability rather than switching to manual control. This dynamic adaptation enables full automation across both known and new scenes, resolving the contradiction between safety and automation extent.
Solution Approach 2:
The system performs self-adaptation by automatically updating model parameters when encountering new scenes, without requiring driver intervention. This self-service capability allows the vehicle to maintain autonomous driving in new scenes, achieving full automation while ensuring safety through adaptive model updating.
3Productivity
If a pre-training model is used for autonomous driving in prescribed scenes, then autonomous driving strategies can be generated, but the model lacks adaptability to handle new and unfamiliar driving environments
Solution Approach 1:
The system transforms the static pre-training model into a dynamic adaptive model by implementing real-time parameter updates based on new scene data. This enables the model to maintain high productivity in generating autonomous driving strategies for prescribed scenes while simultaneously adapting to new and unfamiliar driving environments.
Solution Approach 2:
The system changes model parameters based on scene type detection and new scene data. By updating parameters dynamically, the model maintains efficient strategy generation for known scenes while adapting to new scenarios, resolving the contradiction between productivity and adaptability.
Data Source
AI summary
An adaptive processing method for new scenes in autonomous driving, comprising: obtaining scene data corresponding to new scene of vehicle driving, wherein the scene data describes vehicles state and driving operations in the new scene; obtaining a test set of the new scene based on processing the scene data by a preset distribution; updating parameters of a pre-training model by inputting the test set, and obtaining a scene model adapted to the new scene based on gradient iteration of general model parameters of the pre-training model, wherein the scene model is configured to output an autonomous driving strategy for the vehicle in the new scene. Therefore, the autonomous driving vehicle transforms a new scene to a known scene, and no longer be troubled by unpredictable new scenes, and greatly enhance the reliability and stability of autonomous driving.