Rollover prediction using lateral energy models and active rollover prevention

Jimmy Chiu, Purdue University

Abstract

Each year there are several thousand reported cases of vehicles rolling over in the United States, with 80% of all fatal accidents involving rollover in Sport Utility Vehicles (SUVs), trucks or vans. Cases involving occupants being thrown out from the vehicle during rollover can be up to 10 times more likely to sustain serious injury or death. Determining whether a vehicle will roll over before the event occurs provides advanced warning and the ability to apply active and/or passive safety countermeasures. The objective of rollover mitigation systems can be viewed as a two-part problem: prediction and prevention. Prediction requires a vehicle model that captures the key dynamics during a rollover event. We present a vehicle rollover prediction algorithm that gives the current state as well as future states, allowing the system to determine if the conditions necessary for rollover will be achieved before the conditions are measured by the sensors, increasing lead time to perform preventative countermeasures. We develop a control strategy to prevent rollover utilizing fully active suspension to generate a roll prevention torque on the vehicle. The proposed control law applies a torque to the sprung mass such that the left and right side vertical loads are driven to the nominal load values (under zero lateral acceleration). We consider actuation rate and magnitude limits for such systems based on existing hardware and implement the control on a simulation model in CarSim. Using simulation we demonstrate the prevention of rollover in both the flick turn and fishhook turn maneuvers subject to the actuation constraints. It is shown that this control strategy is able to increase the lateral acceleration limitation before rollover is achieved, verified on highly dynamic maneuvers where the lateral load transfer is maximized. Our goal is to achieve improved roll prevention without requiring the system to be constantly active as such approach has negative effects on the efficiency of the vehicle overall. We show in simulation results that activation of the controller only once liftoff has been predicted increases the maximum sustained lateral accelerations that the vehicle can encounter without rolling over. It is shown with realistic actuation torque limitations that such system can reduce the risk of rollover for highly dynamic steering inputs that maximize lateral load transfer.

Degree

Ph.D.

Advisors

Corless, Purdue University.

Subject Area

Automotive engineering|Mechanical engineering

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server
.

Share

COinS