CAN Message ID Decoding via Power Signal Correlation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current systems fail to provide disaggregated power distribution information within electric vehicles, making it difficult to extract details on driver behavior, vehicle maintenance, and predicted component power output, due to the complexity of CAN-bus messages and lack of disclosed ECU IDs.
Innovation Solution
A method that jointly decodes CAN-bus messages and disaggregates electrical power by using multivariable regression to map CAN-bus message streams to aggregated electrical energy signals, employing heuristics to decode CAN IDs and correlate them with component-level energy usage, thereby enabling comprehensive data storage for component monitoring and optimized power distribution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If CAN-bus messages are used to communicate component information, then data transmission between components is enabled, but the complexity of encoded ECU IDs prevents accurate correlation with vehicle components
Solution Approach 1:
The patent introduces an intermediary decoding system that acts as a mediator between the encoded CAN-bus messages and the component identification requirements. This intermediary layer translates the non-disclosed manufacturer codes into meaningful component identifiers through machine learning models, enabling accurate correlation without modifying the original CAN-bus communication protocol.
Solution Approach 2:
The patent replaces the traditional manual or rule-based ECU ID decoding approach with automated machine learning models. These models learn patterns from aggregated power consumption data and CAN-bus message timing to automatically decode ECU IDs, substituting complex manual decoding mechanics with intelligent automated systems.
2Use of energy by stationary object
If aggregated power consumption data is monitored at the PDU level, then overall electricity usage is tracked, but disaggregation to individual component level is lost
Solution Approach 1:
The patent applies segmentation by dividing the aggregated power consumption signal into individual component-level power signatures. Through machine learning analysis, the system segments the total power draw into distinct contributions from each ECU and component, enabling precise measurement of individual component power usage while maintaining the benefits of centralized PDU monitoring.
Solution Approach 2:
The patent uses partial action by analyzing specific temporal patterns and power signature characteristics of individual components within the aggregated signal. Rather than attempting to measure everything simultaneously, the system focuses on identifying distinctive power consumption patterns of specific ECUs at relevant time intervals, achieving precise component-level measurement through selective analysis.
3Adaptability or versatility
If multiple ECUs are interconnected via CAN-bus, then comprehensive vehicle control is achieved, but the number of undecipherable messages increases
Solution Approach 1:
The patent implements self-service by enabling the system to automatically decode and interpret CAN-bus messages without external intervention. The machine learning models continuously learn from incoming messages and power consumption patterns, allowing the system to self-improve its decoding accuracy over time and handle the increasing complexity of multiple ECUs autonomously.
Solution Approach 2:
The patent applies parameter changes by transforming the decoding approach from static rule-based methods to dynamic machine learning models that adapt their parameters based on learned patterns. The models adjust their interpretation parameters based on timing, power consumption levels, and message frequency, enabling effective handling of increased message complexity from multiple ECUs.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A message ID decoding method for a system comprising multiple components interlinked by a CAN-bus through which messages are sent, the method comprising: acquiring a CAN-bus message data stream comprising multiple CAN-bus messages, each CAN-bus message comprising a CAN ID, each CAN ID indicating a component from which the corresponding CAN-bus message originated, storing the CAN-bus messages in a first datastore, acquiring an electricity consumption signal, wherein the electricity consumption signal is indicative of the electricity consumption of a component in the system, storing the electricity consumption signal in a second datastore, generating time-stamp bins, each time-stamp bin corresponding to an interval in time, selecting, from their respective stores, a CAN ID and a portion of the electricity consumption signal, the CAN ID and the portion of the electricity consumption signal being from an interval in time corresponding to a time-stamp bin, and determining a regression coefficient, R, indicating the level of relatedness between the selected CAN ID and the selected portion of electricity consumption signal, thereby correlating the ID of the selected CAN-bus message with a component.