Browse Prior Art Database

A novel non-intrusive monitor for vehicle rollover detection Disclosure Number: IPCOM000248461D
Publication Date: 2016-Dec-01
Document File: 1 page(s) / 195K

Publishing Venue

The Prior Art Database

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 71% of the total text.

Page 01 of 1

A novel non-intrusive monitor for vehicle rollover detection

Rollover detection is an indispensable safety feature of modern automotive control systems. It detects and predicts rollover events and triggers the deployment of airbags to protect passengers/drivers. We consider the rollover detection algorithm that uses roll rate, lateral acceleration and vertical acceleration to detect rollover events. The present production algorithm consists of a set of calibrations that usually need to be tuned by hand with iterations for about two weeks. Our objective is to develop innovative automated calibration approach to identify those calibrations in a much more efficient manner.


In this work we use support vector machines (SVM) based method to develop a classification model to detect roll over based on measured signals. In an earlier work, SVM was successfully applied to develop a non-intrusive catalyst monitor. SVM is a machine learning method that uses training examples from both the classes (rollover and no rollover) to develop a model that could be used to classify unknown test data. Shown in Fig 1 is the model performance for a vehicle rollover experiment. The red dot represents the nominal or no roll condition while the blue dot represents vehicle roll condition. As one would intuitively expect for higher angle (feature1) and higher angular rate (feature 2) the vehicle would roll. The model is required to find the threshold boundary. As can be seen from Fig...