Predictive Threat Assessment for Early Rollover Detection in Articulated Heavy Vehicles - Detecting Potential Rollovers by Incorporating Road Geometry Data into Predictive Algorithm

Abstract

Abstract Due to the high center of mass and heavy payload, rollovers is a serious threat towards heavy articulated vehicles. Studies have shown that rollovers are more common on poorly designed roads [3, 7], prompting the question of whether future rollover risks can be detected if the road geometry is known. The thesis proposes a predictive threat assessment, that has the potential to detect rollovers up to 3 seconds before they occur, according to the simulation results presented in the manuscript. The predictive threat assessment proposed in this thesis utilizes a linear vehicle model to predict the future state of the vehicle along with uncertainties of the predictions. The linear model was evaluated against a high-fidelity model. To predict the future road properties and the driver inputs, a road model and a predictive driver model was developed. An extended Kalman filter was also implemented using the vehicle model to estimate the vehicles current state and uncertainties. The conclusion from the study is that it is indeed possible to predict the threat of rollovers, given that the road is known. The results sow that the proposed predictive threat assessment algorithm can predict the vehicle state trajectory up to 3s in simulation, while the risk of rollover is quantified as a probability. We show that the proposed method is flexible, where it becomes a question of optimizing tuning parameters to perform an accurate threat assessment.