Control Unit Software Image for Parallel Numerical Simulation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for developing and testing control unit software require multiple physical control units, which is costly and inefficient, as they need to simulate various environments and interactions, limiting the scalability and flexibility of simulations.
Innovation Solution
A computer-implemented method using an artificial neural network or support vector machine to create a software image of a real control unit, trained with supervised or reinforcement learning, allowing for numerical simulation of the control unit's behavior without relying on multiple physical units, enabling parallel simulations and integration of peripheral control units.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If multiple physical control units are used for simulation, then simulation accuracy and realism are improved, but cost and resource consumption increase
Solution Approach 1:
The patent creates a software image (copy) of the control unit's functionality using neural networks or support vector machines. This virtual copy replicates the input-output behavior of the physical control unit, allowing multiple simulations to run using identical software images instead of requiring multiple physical units. The copying principle directly resolves the contradiction by enabling high-fidelity simulations with zero additional hardware.
Solution Approach 2:
The software image serves multiple simulation purposes simultaneously - it can be used for hardware-in-the-loop testing, software-in-the-loop validation, and various environmental scenario testing all with the same virtual control unit model. This universal software-based approach replaces the need for multiple specialized physical control units, reducing both cost and resource requirements while maintaining simulation accuracy.
2Productivity
If multiple physical control units are deployed for parallel simulations, then simulation scalability is improved, but hardware cost and complexity increase
Solution Approach 1:
The software image can be instantiated multiple times in parallel across different computing resources without requiring additional physical control units. Each simulation environment loads the same software image, enabling unlimited parallel simulations constrained only by computational resources rather than hardware availability. This dramatically improves scalability while reducing hardware complexity.
Solution Approach 2:
The patent replaces the mechanical/physical control unit system with a software-based virtual model. This substitution allows parallel simulations to run on standard computing infrastructure rather than requiring multiple specialized hardware devices, thereby improving scalability while eliminating the need for complex hardware configurations and physical device management.
3Measurement precision
If software images are trained with real control unit data, then behavioral accuracy is improved, but data processing time and computational resources increase
Solution Approach 1:
The software image is trained in advance using historical operational data from the real control unit, capturing its behavioral characteristics before deployment. This preliminary training phase, though time-consuming, is performed once during the software image creation process. Once trained, the software image can be deployed immediately for simulations without requiring additional training time, thus resolving the contradiction between achieving high behavioral accuracy and minimizing time loss during actual simulation operations.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer-implemented method (1) for generating a software image (2) suitable for numerical simulation of at least a part of a real control unit (3) is presented and described, wherein the real control unit (3) maps an input vector of control unit input variables (4a) to an output vector of control unit output variables (4b) during operation. The generation of the software image is based on the fact that the software image (2) is generated by an artificial neural network (5) or by a support vector machine with an input vector of image input variables (6a) containing control unit input variables of interest (7a) and with an output vector of image output variables (6b) containing control unit output variables of interest (7b).that the software image (2) is trained (8) by means of supervised learning or reinforcement learning with a plurality of training input vectors (9a) of the control unit input variables (7a) of interest and, in the case of supervised learning, also with a plurality of corresponding training output vectors (9b) of the control unit output variables (7b) of interest, wherein the training input vectors (9a) and the corresponding training output vectors (9b) are obtained from the operation of the real control unit (3), i.e., from the control unit input variables (4a) and the control unit output variables (4b).