CAN Bus Signal Decoding via Supervised Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Decoding messages on a vehicle's CAN bus to extract performance metrics is challenging due to the large number of message identifiers and combinations, making it difficult for engineers to identify relevant signals without extensive manual analysis.
Innovation Solution
A system using supervised machine learning algorithms to isolate message identifiers and byte numbers associated with specific vehicle components by applying testing maneuvers and generating graphs for technician confirmation, with the ability to store scaling factors for performance metric evaluation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If manual analysis of CAN bus messages is performed to identify relevant signals, then decoding accuracy can be achieved, but the time required and complexity of the process increases significantly
Solution Approach 1:
The patent replaces manual mechanical analysis with an automated machine learning system. The ML model automatically identifies relevant signals by analyzing message patterns, byte changes, and correlation with vehicle states, eliminating the need for manual inspection of hundreds of message identifiers and thousands of potential combinations while maintaining high accuracy.
Solution Approach 2:
The patent introduces a machine learning model as an intermediary between the raw CAN bus data and the performance metrics. This intermediary automatically processes the complex message stream, identifies relevant signals through pattern recognition, and outputs decoded performance data, thereby resolving the contradiction between accuracy and time consumption.
2Adaptability or versatility
If comprehensive testing maneuvers are provided to capture all potential signals, then complete performance metric coverage is achieved, but the complexity of the testing process increases
Solution Approach 1:
The patent segments the testing process into manageable components: the system automatically generates targeted testing maneuvers based on the specific component being analyzed, rather than requiring comprehensive testing of all vehicle systems. This segmentation allows the testing process to be adapted to different components (engine, transmission, battery) without increasing overall complexity.
Solution Approach 2:
The patent makes the testing process dynamic and adaptive. The machine learning model analyzes the specific vehicle and component being tested, then automatically generates appropriate testing maneuvers tailored to that context. This dynamic adaptation allows comprehensive coverage for each specific case while keeping the overall process complexity manageable through automation.
3Reliability
If multiple potential message identifiers and byte numbers are identified, then comprehensive signal detection is achieved, but the difficulty of determining the correct signal increases
Solution Approach 1:
The patent implements feedback mechanisms where the machine learning model presents multiple potential signals to the user with confidence scores and visualizations of each signal's behavior during testing maneuvers. The user can provide feedback to confirm or reject signals, and the system learns from this feedback to improve future identifications, thereby reducing the difficulty of determining the correct signal while maintaining comprehensive detection.
Solution Approach 2:
The patent uses visual differentiation (analogous to color changes) to help users distinguish between multiple potential signals. Different signals are presented with distinct visual characteristics in the graphical interface, and the system highlights the most likely correct signal based on analysis confidence, making it easier for users to identify the correct signal among multiple candidates.
Data Source
AI summary
Techniques for identifying certain signals sent over the CAN bus between components of a vehicle are provided herein. Specifically, certain testing maneuvers designed to engage the component of interest are provided to a technician for performing on the vehicle. The messages can be captured from the CAN bus and analyzed, using supervised machine learning algorithms, to isolate the message ids and the byte numbers so that the values of the component of interest may be observed for determining performance metrics. Once identified, these performance metrics may be used to compare with other vehicles or improve the design and performance of the vehicle.