Sensor Fusion Motion State Determination for Autonomous Vehicles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current autonomous driving systems face challenges in accurately determining the motion state of target vehicles, particularly in integrating data from camera and radar sensors to enhance safety and navigation, especially at higher levels of automation where precise object tracking and motion prediction are crucial.
Innovation Solution
The system employs a combination of camera and radar sensors to detect and classify objects, utilizing high-definition maps for enhanced accuracy, and implements a sensor fusion architecture that includes a dynamic object tracking module using an Extended Kalman Filter (EKF) to track objects' six-degrees-of-freedom motion, including position, orientation, and velocity, while also leveraging machine learning for improved object classification.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If multiple sensors (camera and radar) are integrated to determine motion state, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent combines camera and radar sensors into a unified sensor fusion architecture that processes data from both sources simultaneously. The system merges image data from the camera with range and velocity data from the radar to determine comprehensive motion states of target vehicles, achieving higher measurement precision through multi-sensor integration.
Solution Approach 2:
The sensor fusion architecture is designed to handle multiple functions: object detection, classification, tracking, and motion state determination. The system uses a unified processing framework that can process different sensor types (camera images and radar data) through common modules including the Extended Kalman Filter for tracking and machine learning classifiers for object identification.
2Measurement precision
If sensor fusion with Extended Kalman Filter is used for six-degrees-of-freedom tracking, then object tracking precision is improved, but computational requirements increase
Solution Approach 1:
The tracking system segments the six-degrees-of-freedom motion parameters into distinct components: position (x, y, z coordinates) and orientation (roll, pitch, yaw angles). The Extended Kalman Filter processes these segmented parameters separately, allowing for more efficient computational handling while maintaining high tracking precision for each degree of freedom.
Solution Approach 2:
The system performs preliminary classification of detected objects using machine learning algorithms before initiating full six-degrees-of-freedom tracking with the Extended Kalman Filter. This preliminary action filters out non-relevant objects, reducing the computational burden on the tracking system while maintaining high precision for tracked targets.
3Measurement precision
If machine learning is applied for object classification, then classification accuracy is improved, but processing time increases
Solution Approach 1:
The system performs preliminary processing of sensor data including object detection and feature extraction before applying machine learning classification algorithms. This preliminary action prepares the data in an optimized format, allowing the machine learning models to process information more efficiently and reduce overall processing time while maintaining high classification accuracy.
Solution Approach 2:
The system uses pre-trained machine learning models that have been previously trained on large datasets of objects and scenarios. These pre-trained models (copies of optimized classifiers) can be deployed in the autonomous vehicle system without requiring real-time training, significantly reducing processing time while maintaining high classification accuracy through the use of established model architectures.
Data Source
AI summary
Disclosed are techniques for determining a motion state of a target object. In an aspect, an on-board computer of an ego vehicle detects the target object in one or more images, determines one or more first attributes of the target object based on measurements of the one or more images, determines one or more second attributes of the target object based on measurements of a map of a roadway on which the target object is travelling, and determines the motion state of the target object based on the one or more first attributes and the one or more second attributes of the target object.