ABS Controller Modeling via Deep Learning Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The existing integrated virtual vehicle performance analysis systems face difficulties in modeling and testing anti-lock brake system (ABS) controllers due to proprietary algorithms, leading to increased development costs and time, especially with the integration of advanced driver assistance systems and autonomous driving, as public access to internal logic is limited.
Innovation Solution
A deep learning-based system using machine learning networks, comprising a vehicle speed estimation network, wheel speed state estimation network, and classification network, which estimates vehicle speed, wheel speed characteristics, and braking modes using wheel speed data and brake pedal information, allowing for accurate modeling of ABS controllers without requiring additional data processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning networks are used to model ABS controllers, then modeling accuracy is improved, but device complexity increases
Solution Approach 1:
The deep learning model is divided into three separate networks: vehicle speed estimation network, wheel speed state estimation network, and classification network. Each network performs a specific function, making the overall complex system manageable and modular while maintaining high modeling accuracy.
Solution Approach 2:
The patent introduces intermediate estimation steps (vehicle speed estimation and wheel speed state estimation) as mediators between the input wheel speed data and the final braking mode classification. This intermediary approach improves accuracy by breaking down the complex mapping relationship into manageable estimation stages.
2Reliability
If actual vehicle driving tests are performed to evaluate ABS performance, then reliability of performance evaluation is improved, but loss of time increases
Solution Approach 1:
The patent creates a virtual copy of the ABS controller using deep learning networks that replicates the behavior of the actual controller. This virtual model allows performance evaluation in a simulated environment, maintaining reliability while eliminating the need for repeated actual vehicle driving tests.
Solution Approach 2:
The deep learning model is trained in advance using wheel speed data to predict braking modes before actual performance testing. This preliminary modeling action allows developers to evaluate performance scenarios virtually, reducing the need for time-consuming actual vehicle tests.
3Measurement precision
If comprehensive data processing is performed to improve modeling accuracy, then measurement precision is improved, but device complexity increases
Solution Approach 1:
The patent extracts and processes only the essential features from wheel speed data: vehicle speed information and wheel speed state (increasing/decreasing) information. By taking out only the necessary features rather than processing all possible data, the system maintains high accuracy while avoiding unnecessary complexity.
Data Source
AI summary
A system of modeling an anti-lock brake system (ABS) controller of a vehicle includes: a vehicle speed estimation network configured to estimate a vehicle speed through machine learning using wheel speed data of each of a plurality of wheels of the vehicle; a wheel speed state estimation network configured to estimate a time series characteristic of the wheel speed through machine learning using information on the wheel speed data and information on whether a brake pedal is depressed; and a classification network configured to estimate a braking mode for controlling an increase, a decrease, or steady state of a braking pressure of each wheel through machine learning using speed estimation data of the vehicle estimated by the vehicle speed estimation network and time series characteristic estimation data of the wheel speed estimated by the wheel speed state estimation network.