Vehicle Sensor Data Compression for Object Detection Cost Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for processing sensor data in control devices, such as for automated driving, result in high storage, transmission, and processing costs due to limited data reduction using lossless compression, typically reducing data volume by only 40% to 60%.
Innovation Solution
Implementing lossy compression in control devices to irreversibly remove irrelevant information from sensor data, reducing data volume by at least 40% to 80%, and evaluating the impact of lossy compression on object detection algorithms without additional comparison data, using a control device with integrated hardware modules for compression and decompression.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If lossless compression is used to reduce data volume, then data storage and transmission costs are reduced, but data volume reduction is limited to only 40% to 60%
Solution Approach 1:
The patent changes the compression parameter from lossless to lossy compression, accepting controlled information loss in exchange for significantly higher compression ratios (at least 40%, preferably 60% or more). This parameter change enables the control device to achieve substantial data volume reduction while maintaining sufficient detection performance through optimized compression algorithms that preserve critical features for object detection.
2Quantity of substance
If lossy compression is applied to reduce data volume by at least 40%, then storage and transmission costs decrease, but information is irretrievably removed from the sensor data
Solution Approach 1:
The lossy compression algorithm extracts and removes irrelevant or redundant information from sensor data while retaining critical features necessary for object detection. By selectively removing non-essential data components, the system achieves significant compression ratios while preserving the information needed for accurate detection algorithms.
Solution Approach 2:
The compression algorithm adjusts parameters to optimize the balance between compression ratio and information retention, removing only the least important information while preserving features critical for detection tasks.
3Reliability
If large volumes of uncompressed sensor data are stored and transmitted for algorithm testing, then comprehensive training and validation can be performed, but costs for storage, transmission and processing increase significantly
Solution Approach 1:
The system changes the data processing parameter by applying lossy compression before storage and transmission, then decompressing for algorithm testing. This parameter change maintains detection performance validation accuracy while dramatically reducing storage and transmission costs, as the compressed data can be efficiently stored and transmitted then fully restored for comprehensive algorithm testing.
4Manufacturing precision
If lossless compression is used to maintain data integrity, then exact reconstruction is possible, but data volume reduction is insufficient to significantly lower costs
Solution Approach 1:
The patent transitions from lossless to lossy compression parameters, accepting controlled information loss to achieve substantially higher data volume reduction. This parameter change enables cost-effective deployment by reducing storage and transmission requirements by at least 40%, while detection performance remains sufficient for practical applications.
Data Source
AI summary
A method for processing sensor data in a control device using lossy compression. The method includes: receiving the sensor data in the control device, wherein the sensor data have been provided by a sensor system for capturing an environment of a vehicle; compressing the sensor data in a lossy compression method in order to obtain compressed sensor data; decompressing the compressed sensor data in order to obtain decompressed sensor data; and inputting the decompressed sensor data as input data into an object detection algorithm, which is configured to convert the input data into output data which display objects in the environment of the vehicle.