Vehicle Environmental Model Updates for Low-Confidence Dynamic Objects
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous vehicles face challenges in identifying and classifying objects, leading to uncertainties in environmental models, especially when sensor malfunctions or occlusions occur, requiring efficient communication of specific information to reduce data overload and enhance object recognition confidence.
Innovation Solution
A method for updating environmental models by assigning correctness probabilities to dynamic objects and transmitting broadcast messages to request further information only on objects with low confidence, allowing other vehicles or infrastructure to provide additional data, thereby improving object classification and reducing communication overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor data is shared between vehicles to improve environmental model accuracy, then object recognition confidence increases, but communication channel overload occurs
Solution Approach 1:
The patent extracts only the essential uncertainty information (correctness probabilities below threshold) from the complete sensor data, transmitting only this critical subset rather than full environmental models. This extraction principle reduces communication overhead while maintaining the ability to improve object recognition confidence through targeted information sharing.
Solution Approach 2:
The patent applies local quality by focusing communication resources on specific objects with low correctness probabilities rather than uniformly processing all detected objects. Each object receives differential treatment based on its uncertainty level, with high-uncertainty objects receiving additional verification requests and high-confidence objects being efficiently excluded from further communication.
2Measurement precision
If complete environmental models are transmitted between vehicles, then object classification accuracy improves, but communication efficiency decreases
Solution Approach 1:
The patent extracts only the critical uncertainty information (correctness probabilities below threshold) from the complete environmental model, transmitting this condensed subset between vehicles. This extraction maintains the ability to improve object classification accuracy through selective information sharing while dramatically reducing communication data volume and improving efficiency.
Solution Approach 2:
The patent applies partial action by transmitting only the necessary portion of environmental model information (objects with low correctness probabilities) rather than complete models. This partial information exchange is sufficient to improve classification accuracy for uncertain objects without the overhead of complete model transmission.
3Reliability
If all detected objects are communicated for verification, then environmental model certainty increases, but communication overhead increases
Solution Approach 1:
The patent applies local quality by differentiating communication treatment based on object uncertainty levels. Objects with correctness probabilities below the threshold receive verification requests to improve environmental model certainty, while objects with high confidence are excluded from communication. This localized approach achieves reliability improvement without proportional increases in communication data volume.
Solution Approach 2:
The patent applies partial action by performing verification requests only on the necessary subset of objects (those with low correctness probabilities) rather than all detected objects. This partial verification approach achieves sufficient environmental model certainty for safe autonomous operation without the excessive communication overhead of complete model verification.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Embodiments provide methods, computer programs, apparatuses, a vehicle, and a traffic entity for updating an environmental model at a vehicle. The method (10) for a vehicle (100) and for updating an environmental model of the vehicle (100) comprises obtaining (12) an environmental model of the vehicle (100), the environmental model comprising static and dynamic objects in the environment of the vehicle (100) along at least a part of the vehicle's trajectory. The method (10) further comprises assigning (14) information related to correctness probabilities at least to dynamic objects of the environmental model, and determining (16) at least one dynamic object for which the information related to the correctness probability indicates a correctness probability below a threshold. The method (10) further comprises transmitting (18) a broadcast message to the environment to request further information on the at least one dynamic object.