The rising popularity of Sport Utility Vehicles (SUV s) and other vehicles with high centers of mass has heightened the importance of effective anti-rollover protection systems. Traditional methods for checking roll rates can be cost-prohibitive due to the need for redundant sensors. This paper presents a novel approach using a Long Short-Term Memory (LSTM) neural network to predict the roll rate of vehicles under off-road conditions, forecasting up to 0.04 seconds into the future. The model inputs include lateral and longitudinal accelerations, roll and yaw rates, the steering angle of the first axle, and the speeds of all four wheels, with data spanning the current and up to 0.04 seconds in the past. The neural network was trained and validated with five hours of preprocessed data from the Bremly, Michigan off-road test track. A comparison with results from existing literature, along with a thorough evaluation, robustly demonstrates the effectiveness and reliability of the proposed model in predicting vehicle roll rates.
Forecasting Roll Rate for Sensor Plausibility Check in Off-Road Vehicle Conditions
2024-08-26
1529849 byte
Conference paper
Electronic Resource
English
Roll rate plausibility checking and estimation of vehicle attitude using a sensor signal platform
Automotive engineering | 2007
|