A Parallel Optimization Scheme for Parameter Estimation in Motor Vehicle Dynamics

摘要

For calibrating the vehicle model of a commercial vehicle dynamics program a parameter estimation tool has been developed which relies on observations obtained from driving tests. The associated nonlinear least-squares problem can be solved by means of mathematical optimization algorithms most of them making use of first-order derivative information. While the complexity of the investigated vehicle dynamics program only allows the objective gradients to be approximated by means of finite differences, this approach enables significant savings in computational time when performing the additionally required evaluations of the objective function in parallel. The employed low-cost parallel computing platform which consists of a heterogeneous PC cluster is well suited for the needs of the automotive suppliers and industries employing vehicle dynamics simulations.