Blind Spot Obstacle Attribute Estimation via V2X Data Fusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Autonomous driving vehicles face challenges in accurately determining the attribute values of obstacles, particularly those in blind spots, due to limited sensor coverage and incomplete data sets, which can lead to delayed or inaccurate decision-making.
Innovation Solution
The method involves acquiring vehicle-end data from sensors and V2X data from roadside devices, and fusing this information to estimate attribute values of obstacles at the edge of blind spots using data fusion techniques, such as Kalman filters and hidden Markov models, to enhance accuracy and timeliness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensor coverage is increased to detect obstacles in blind spots, then detection coverage is improved, but device complexity and cost increase
Solution Approach 1:
The patent introduces roadside devices as intermediary elements that act as external sensors positioned at strategic locations around the vehicle's blind spots. These roadside devices detect obstacles in areas where vehicle-mounted sensors cannot effectively observe, thereby extending detection coverage without adding complex sensors to the vehicle itself. The detection data from these intermediary roadside devices is then integrated into the vehicle's obstacle detection system.
2Measurement precision
If data fusion techniques are applied to combine vehicle-end and V2X data, then attribute estimation accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent applies preliminary action by pre-training deep learning models offline before deployment. The attribute estimation model is trained in advance using large datasets to learn the complex relationships between sensor data and obstacle attributes. During actual operation, the pre-trained model performs inference with reduced computational burden, as the heavy learning process has already been completed beforehand. This allows accurate attribute estimation without requiring excessive real-time computational resources.
3Measurement precision
If deep learning models are used for attribute estimation, then estimation accuracy is improved, but processing time and computational resources increase
Solution Approach 1:
The patent implements preliminary action through offline pre-training of deep learning models. The complex attribute estimation model is trained in advance using comprehensive datasets, allowing it to capture intricate patterns and relationships. During real-time operation, the pre-trained model performs rapid inference with minimal processing time, as the computationally intensive learning phase has already been completed beforehand. This separation of training and inference phases enables high accuracy without real-time computational delays.
Solution Approach 2:
The patent employs dynamic adaptation by continuously updating and optimizing the deep learning model based on incoming data streams and feedback. The model can adapt its parameters and structure dynamically during operation to maintain high accuracy while optimizing processing efficiency. This dynamic adjustment allows the system to balance between estimation accuracy and processing speed according to real-time requirements.
Data Source
AI summary
The present disclosure provides a method and apparatus for determining an attribute value of an obstacle in vehicle infrastructure cooperation. The method includes: acquiring vehicle-end data collected by at least one sensor of an autonomous driving vehicle; acquiring vehicle wireless communication V2X data transmitted by a roadside device; and fusing, in response to determining that an obstacle is at an edge of a blind spot of the autonomous driving vehicle, the vehicle-end data and the V2X data to obtain an attribute estimated value of the obstacle.