Vehicle Sensor Fusion Using Mixed Raw and Target Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge of efficiently processing and fusing large amounts of data from multiple vehicle sensors in autonomous driving systems leads to high overheads and potential loss of target information.
Innovation Solution
A vehicle sensor data processing system that configures some sensors to output raw data and others to output processed target information, using fusion modules to combine this data and reduce transmission and processing overheads while ensuring accurate fusion and sensing performance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If all sensors output raw data for fusion, then fusion accuracy is improved, but data transmission amount and processing overhead increase significantly
Solution Approach 1:
The patent segments sensors into two categories: first sensors that output raw data and second sensors that output processed target information. This segmentation allows the system to selectively transmit only necessary processed data from certain sensors, reducing overall data transmission volume while maintaining fusion accuracy through selective use of raw data from first sensors.
Solution Approach 2:
Second sensors perform preliminary processing of raw data to generate target information (such as object detection results, tracking data, or extracted features) before transmission. This preliminary action reduces the data transmission amount by pre-processing data at the sensor level, eliminating the need to transmit all raw data to the fusion module.
2Loss of information
If all sensors output raw data for fusion, then complete information is preserved, but processing overhead and system complexity increase
Solution Approach 1:
The fusion module is segmented to handle different data types differently. It processes raw data from first sensors completely while processing target information from second sensors through specialized routines that leverage the pre-processed nature of this data, reducing overall processing overhead while maintaining information completeness.
Solution Approach 2:
Target information from second sensors acts as an intermediary that bridges raw sensor data and final fusion results. This intermediary data format reduces processing complexity by providing pre-extracted meaningful information that requires less computational effort to integrate into the final fusion output.
3Quantity of substance
If second sensors process raw data to output target information, then data transmission amount is reduced, but risk of target information loss increases
Solution Approach 1:
The patent segments the sensor system into first sensors that output raw data and second sensors that output processed target information. This segmentation ensures that critical sensors (first sensors) transmit complete raw data without processing, thereby preserving all target information, while other sensors (second sensors) can reduce data transmission by outputting processed information.
4Reliability
If a large number of sensors are deployed for autonomous driving, then sensing performance is improved, but data fusion complexity increases
Solution Approach 1:
The patent segments the sensor system into first sensors outputting raw data and second sensors outputting processed target information. This segmentation strategy manages data fusion complexity by allowing the fusion module to selectively process data from different sensor types according to their output formats, making the fusion process more scalable and manageable as sensor数量 increases.
Data Source
AI summary
Vehicle sensor data processing methods, apparatuses and system are provided. In an example method, some sensors are configured to output sensed raw data, and other sensors are configured to output target information determined based on sensed raw data. A fusion component performs data fusion after receiving the raw data and the target information from the sensors, to determine target information.