Lorenz Georg Görne presents a method (PrOComp) for optimal usage of the transmission path between the vehicle and the data backend. The compression ratio of vehicle measurement data could be improved from roughly a factor of ten in conventional methods, to up to 27. The method allows vehicle measurement data to be transmitted optimally in terms of data volume via the mobile internet and via traditional transmission routes. Through the PrOComp method, real-time data analysis over the mobile internet is feasible, as well as the collection of big data in the field. This enables key features like predictive maintenance, reactive event evaluation (for example crash events) or fast generation of AI training data. Through the usage of standardized interfaces and data formats, PrOComp can be adapted to the needs of many industry branches that feature field data collection.
Author(s): Lorenz Georg Görne
Publisher: Springer
Year: 2023
Language: English
Pages: 166
Foreword
Contents
List of Figures
List of Tables
Acronyms
Symbols
Kurzfassung
Abstract
1 Introduction and Motivation for Mobile Data Transmission
1.1 Main Components for Vehicle Data Transmission
1.1.1 Vehicle Data Collection
1.1.2 Data Filters / Processing
1.1.3 Transmission from Vehicle to Server
1.2 Challenges of Vehicle Data Transfer
1.3 Goal of the Thesis
1.4 Structure of this Work
2 Foundation and State-of-the-Art
2.1 Data Generation within the Vehicle
2.1.1 Vehicle Data Busses and Vehicle Bus Networks
2.1.2 Representation of Vehicle Signals on Bus Systems
2.2 State-of-the-Art Vehicle Data Collection
2.2.1 Common Bus Communication Recording Formats
2.2.2 Formats to Describe Vehicle Signals
2.3 Data Transmission over Mobile Internet
2.3.1 Components of Mobile Internet Transmission
2.3.2 HTTP as Basis of Possible Data Transmission Protocols
2.3.3 HTTP-based Data Transmission Protocols
2.4 Data Processing Prior to Data Transfer
2.4.1 Data Reduction through Filters or Aggregation
2.4.2 Compression Theory
2.4.3 General Purpose Compression Algorithms
2.4.4 Compression Algorithms Optimized for Vehicle Bus Data
3 Model of the Transmission Pipeline
3.1 Model of Vehicle Communication Interface
3.2 Model of Data Pre-Processing
3.2.1 Modeling of CPU Tasks
3.2.2 Introduction of Data Blocks within a Pipeline
3.2.3 Estimation of Pre-Processing Throughput and Buffer Size
3.3 Model of Data Compression
3.3.1 Main Effects of Data Compression
3.3.2 Curve Fitting for Available Compression Algorithms
3.4 Model of Data Transfer
3.4.1 Observable Factors for Internet Data Transmission
3.4.2 Introduction of Net and Gross Network Speed
3.4.3 Effect of Transmission Control Protocol/Internet Protocol (TCP/IP)
Protocol Overhead
3.4.4 Additional Overhead for High-Level Protocols
3.4.5 Effect of Connection Interruptions and Outages on Transmission Speed
3.4.6 Conclusion for Data Transfer Modeling
3.5 Model of Factors Subject to Optimization
3.5.1 Model of Intermediate Data Buffers
3.5.2 Model of System Latency
3.5.3 Conclusion of the Optimization Model
4 Introduction of the PrOComp Method
4.1 Data Flow Architecture of the Processing Pipeline
4.2 Specification for the Pre-Processing Techniques
4.2.1 1st: Bus Signal Transformation (BST)
4.2.2 2nd: Differential Signal Transformation (DST)
4.2.3 3rd: Signal Alignment Transformation (SAT)
4.2.4 Pre-Processing Results Data Format: Compression Friendly Format
(CFF)
4.2.5 Program Flow of the Pre-Processing Step
4.2.6 Algorithmic Complexity of the Pre-Processing Step
4.3 Specification for the Compression Step
4.3.1 Compression Algorithms Considered in thisWork
4.3.2 Parameters of the Compression Step
4.4 Specification for Data Transfer
4.4.1 Choice of a Suitable Transmission Protocol
4.4.2 Suggested HTTP Protocol for Data Transmission
4.4.3 Additional Considerations and Requirements for Application
4.5 Observation of Key Pipeline Factors
4.5.1 Preliminary Measurements for Pipeline Calibration
4.5.2 Runtime Observation of Pipeline Factors
4.6 Parameter Optimization
4.6.1 Goals for Parameter Optimization
4.6.2 1st Strategy: Maximization of CR
4.6.3 2nd Strategy: Minimization of Preliminary Buffers
4.6.4 3rd Strategy: Minimization of Overall Latency
4.6.5 Runtime Adaption to Model Inaccuracies
4.6.6 Interpolation of Discrete Compression Algorithms
5 Measurement Results and Validation of the Method
5.1 Description of Vehicle Data Collection and Measurements
5.1.1 Vehicle Data Measurement Results
5.1.2 Conclusions about Vehicle Data Measurements
5.2 Performance Measurements of Processing Pipeline Steps
5.2.1 Pre-Processing Step Performance
5.2.2 Compression Step Performance
5.2.3 Curve Fitting on Compression Algorithms
5.3 Evaluation of Factors on CR
5.3.1 Effect of Variation of the Vehicle Data Bus
5.3.2 Effect of the Vehicle State on CR
5.3.3 Effect of the Process Parameter Sblock on CR
5.4 Optimization Strategy Evaluation Based on the Measurements
5.4.1 Evaluation of the Buffer Size Minimization Strategy
5.4.2 Evaluation of the Latency Minimization Strategy
5.5 Conclusion and Validation of the PrOComp method
5.5.1 Improvements of PrOComp over State-of-the-Art
5.5.2 Room for CR Improvements
6 Discussion and Outlook
6.1 Critical Analysis of the Method
6.1.1 Possibilities to Improve the Compression Efficiency
6.1.2 Necessary Measures for Transmission of Signal Data over Mobile
Internet
6.2 Relevance of this Work and Outlook
Bibliography
Appendix
A.1 Comparison of Compression Schemes
A.2 Detailed Description of the CFF Data Format
A.3 Transformation Techniques Description and Complexity Analysis
A.4 Overview of Data Base Can (DBC) Files Used in the Vehicle Measurements
A.5 Effect of Sblock on the overall CR
A.6 Pre-Processing Techniques and Their Individual Effect on
CR
