Vehicle Load Estimation Using Recursive Total Least Squares for Rollover Detection

This paper will describe the development of a load estimation algorithm that is used to estimate the load parameters necessary to detect a vehicle’s proximity to rollover. When operating a vehicle near its handling limits or with large loads, vehicle rollover must be considered for safe operation. Vehicle mass and center of gravity (CG) height play a large role in a vehicle’s rollover propensity. Cargo and passenger vehicles operate under a range of load configurations; therefore, changes in load should be estimated. Researchers have often developed load estimation and rollover detection algorithms separately. This paper will develop a load estimation algorithm and use the load estimates and vehicle states to detect rollover. The load estimation algorithm uses total least squares and is broken into two parts. First, mass is estimated based on a “full-car” dynamic ride model. Next, the CG height and inertia are estimated using the previously estimated mass and a dynamic roll model. Least squares is a popular method for load estimation. Least Squares (LS) assumes that there is no measurement noise which is violated in this application. Total Least Squares (TLS) accounts for measurement noise and provides more accurate estimates when measurement noise is present. Simulated data from CarSim is used to produce sensor measurements. Inertial measurement unit (IMU) and suspension defection sensors are used to measure the appropriate vehicle states. Noise is added to each measurement. Accuracy of the load estimation will be discussed and compared to the least squares approach. Rollover detection using load estimates will be analyzed and compared to rollover detection that does not account for changes in load.