Lessons Learned From Matching Experimental Data to Low-Order Models of Vehicle Behavior

Abstract

This work presents “lessons learned” from an ongoing experimental and simulation investigation of vehicle chassis dynamics. The overall goal of this work is to find low-order physics-based models that are easily fitted from experimental data, yet accurately describe vehicle chassis motion in yaw, sideslip, and roll angle. In previous work related to this effort [5], anomalies in the model were found where key tire parameters, specifically the cornering stiffnesses, exhibit different best-fit values depending on whether the vehicle is tested in swept sine frequency responses or curvilinear steadystate motion such as skid pad maneuvers. The first portion of this work summarizes these results. Next, new research is presented investigating the source of cornering stiffness discrepancies. The tire modeling errors are found to be very strongly related to vehicle roll angle, and from this insight, corrections to the simple chassis models are derived to allow inclusion of a roll dependency in the tire model. Additionally, during investigation of this advanced tire model, terrain disturbances were found to be significant. A method to remove terrain effects is proposed and demonstrated. Comparing these roll- and terrain-corrected fits to results from prior work [5], the new model gives a significantly improved fit in the frequency domain, and outstanding fit in the time domain.