Autonomous Vehicle Object Classification via User Feedback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current automated vehicle systems face challenges in accurately classifying objects as obstacles or not, leading to potential incorrect trajectory planning and navigation issues, especially with unknown or misclassified objects like leaves or cardboard boxes, which can cause vehicles to become stuck.
Innovation Solution
A method where camera images of unclassifiable objects are transmitted to a user's mobile communication device for classification, allowing the user to provide classification information that is then used in the vehicle's trajectory planning, leveraging the user's assessment to overcome incorrect classifications and enable safe navigation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Extent of automation
If automated classification algorithms are used to classify objects as obstacles or not, then the vehicle can operate autonomously without continuous human input, but the classification accuracy decreases leading to incorrect trajectory planning
Solution Approach 1:
The patent introduces a mobile communication device as an intermediary between the automated classification system and the user. When the classifier is uncertain about an object's classification, the system transmits images of the object to the user's mobile device, allowing the user to provide clarification without requiring continuous manual monitoring of the automated driving system.
Solution Approach 2:
The system implements a feedback mechanism where classification results are evaluated for certainty, and when uncertainty is detected, additional information (images of the object) is transmitted to the user for feedback. This feedback loop allows the system to resolve ambiguous classifications while maintaining automated operation.
2Measurement precision
If the vehicle transmits object images to the user's mobile device for classification, then the classification accuracy improves, but the system complexity and communication requirements increase
Solution Approach 1:
The system uses the user's existing mobile communication device, which they already carry and are familiar with. This eliminates the need for the vehicle to provide a complex display or input interface, as the user's personal device serves the classification function.
Solution Approach 2:
The mobile communication device acts as an intermediary that leverages existing user technology rather than requiring the vehicle to implement a complex user interface system. This reduces the vehicle's system complexity while maintaining the ability to obtain accurate classifications.
3Reliability
If the vehicle requests user input for object classification, then the reliability of trajectory planning improves, but the operation time increases due to communication delays
Solution Approach 1:
The system does not require user input for all objects, but only transmits images for objects where the classifier is uncertain. This partial action approach maintains automated operation for clear cases while seeking user input only when necessary, minimizing time loss.
Solution Approach 2:
The system performs preliminary classification using automated algorithms before seeking user input. This preliminary action filters out most objects that don't require user attention, so communication with the user is needed only for ambiguous cases.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
The invention relates to a method for operating an automatically driven, driverless motor vehicle (1), in particular a passenger vehicle, in which sensor data is evaluated with respect to the objects (17) to be taken into account during the route planning, said data being captured by the ambient sensors of the motor vehicle (1) which comprise at least one camera (6), said objects (17) can be classified according to at least one classifier evaluating the associated sensor data as an obstacle or no obstacle. When the object (17) can not be classified, or it can not be classified with sufficient reliability as an obstacle or not an obstacle and/or in the presence of at least one object (17) obstructing the continuation of a route of the motor vehicle (1) to a current target destination, at least one camera image (18) of the corresponding object (17) is captured by at least one of the at least one cameras (6), is transmitted to a mobile communication device (11) carried by a user (12) of the motor vehicle (1) and is represented on said device, and an input of the user (12) classifying the object (17) as an obstacle or not an obstacle is captured as classifying information, said classifying information being sent back to the motor vehicle (1) and is taken into account for the further automatic driving of the motor vehicle (1).