Warning and control for vehicle rollover prevention.

Abstract

This dissertation focuses on the development of a Time-To-Rollover (TTR) metric that can accurately assess the rollover threat under a wide range of vehicle speeds and steering patterns. There are two conflicting requirements to implement TTR in real-time. On one hand, a faster-than-real-time vehicle model is needed. On the other hand, the TTR predicted by this model needs to be accurate enough under all driving scenarios. An innovative approach using hybrid models-Neural Networks (NN) is proposed to solve this dilemma. Two case studies are presented in this dissertation. In the first case study, the TTR metric was utilized as the basis of rollover warning for an articulated heavy truck. A simple decoupled yaw-roll model was developed for TTR calculation. A NN was then developed and trained to mitigate the accuracy problem of this simple model. The NN-TTR metric was found to be accurate across an array of test scenarios. In the second case study, the TTR metric is utilized as the basis of rollover warning and anti-rollover control for sport utility vehicles. The test data of a 1997 Jeep Cherokee was used to construct a TruckSim model and to verify the TTR and NN-TTR metrics. The TruckSim model was used to verify the anti-rollover control algorithm and to simulate the vehicle dynamics in the UM-Oakland driving simulator. The proposed control algorithm was compared with other threshold-based control algorithms in TruckSim. A human-in-the-loop experiment was then conducted to study the performance of the proposed control algorithm by using the driving simulator. The first control algorithm we tested was found to perform unsatisfactorily. It was then redesigned by using direct yaw moment control and the control gain was optimized by using the UMTRI driver model. The redesigned control showed significant improvement for the new human-in-the-loop experiment. The robustness of the TTR metric was studied against selected variations and uncertainties. The TTR metric was found to be robust against vehicle load variation (for SUVs). With some extra measurements, the robustness against low tire pressure, superelevation, and measurement noises were found to be acceptable.