Autonomous Driving Assistance With Confidence-Triggered Remote Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing autonomous driving devices fail to continue performing tasks in complex terrain or road conditions due to preset autonomous driving models' inability to process scene information effectively.
Innovation Solution
An autonomous driving assistance method that processes scene information using a preset model, determines confidence levels, and initiates assistance requests to remote interactive terminals to receive and execute driving instructions, optimizing the model with user-triggered feedback to enhance processing capabilities.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If a preset autonomous driving model is used to process scene information, then the autonomous driving task can be performed initially, but the system fails when encountering complex terrain or road conditions that the model cannot handle
Solution Approach 1:
The system implements a feedback mechanism where processing results with confidence below the threshold trigger assistance requests. The assistance device returns optimized driving instructions that feed back into the system, enabling the preset model to handle complex scenarios it initially cannot process, thus maintaining task continuity while improving adaptability.
Solution Approach 2:
An assistance device acts as an intermediary between the preset autonomous driving model and the actual driving task execution. When the model encounters complex terrain or road conditions it cannot handle, the assistance device receives the request, processes the scene information, and returns optimized driving instructions, effectively mediating the gap between model capabilities and task requirements.
2Extent of automation
If the autonomous driving model is preset in the driving device, then the device can operate autonomously, but it cannot adapt to new or complex scenarios without external assistance
Solution Approach 1:
The system dynamically adjusts its operation mode based on scene complexity. When confidence in the preset model's processing result exceeds the threshold, the system operates fully autonomously. When confidence falls below the threshold, it dynamically switches to requesting assistance from the assistance device, thus maintaining high automation while adapting to scenario complexity.
Solution Approach 2:
The system uses confidence threshold feedback to determine when to operate autonomously and when to seek external assistance. This feedback mechanism allows the preset model to maintain autonomous operation for familiar scenarios while automatically recognizing when external help is needed for complex scenarios, balancing automation with adaptability.
3Adaptability or versatility
If the system requests assistance for every uncertain scenario, then adaptability improves, but system complexity and communication overhead increase
Solution Approach 1:
The system changes the parameter of confidence threshold to control when assistance requests are triggered. By adjusting this threshold parameter, the system balances between maintaining simplicity (higher threshold, fewer requests) and improving adaptability (lower threshold, more requests), thus managing the trade-off between system complexity and scenario handling capability.
Solution Approach 2:
Instead of requesting assistance for every uncertain scenario, the system applies partial action by only requesting assistance when confidence falls below the threshold. This selective approach avoids the excessive complexity of universal assistance requests while still maintaining adequate adaptability for truly complex scenarios.
Data Source
Figure 1~2
Figure 3
Figure 4
AI summary
An autonomous driving assistance method, a driving device, an assistance device and a readable storage medium are provided by the present invention. The driving device processes collected current scene information by using a current autonomous driving model, and initiates an assistance driving request to an assistance device according to a processing result; receives and performs a driving instruction fedback by the assistance device, where the driving instruction is used to optimize the current autonomous driving model in conjunction with the current scene information, so as to perform a following autonomous driving task by using the optimized autonomous driving model. Therefore, a problem, that when performing an autonomous driving task, an existing deriving device fails to continue performing the autonomous driving task since a preset autonomous driving model cannot process scene information, can be solved, improving intelligence and applicability of the autonomous driving model.