Using High-Fidelity Multibody Vehicle Models in Real-Time Simulations 2012-01-0927
Digital or virtual prototyping by means of a multibody simulation model (MBS) is a standard part of the automotive design process. A high-fidelity model is built and often correlated against test data to increase its accuracy. Once built the MBS model can then be used for high fidelity analysis in ride comfort, handling as well as durability. Next to the MBS model, current industry practice is to develop a reduced degree of freedom model for the design and validation of control or intelligent systems. The models used in the control system design are required to execute in hardware-in-the-loop (HIL) simulations where it is necessary to run real-time. The reason for the creation of the reduced degree of freedom models so far has been that the high-fidelity or off-line model does not execute fast enough to be used in an HIL simulation. A HIL simulation requires the model to be simplified by removing bodies and degrees of freedom until the model has been reduced enough to execute real-time. Typically, real-time vehicle models contain about 15 independent degrees of freedom while the MBS model often contain 50 or more degrees of freedom. The large number of degrees of freedom in the MBS model arises from the suspension bodies and by including compliant elements and not just kinematic constraints between the suspension bodies. This reduction process weakens the connection between the MBS model and the real-time model with the result that parameters in the MBS model are not directly available in the real-time model.
Therefore, to maintain a tight connection between the MBS and real-time model it is necessary to be able to execute the MBS model in real-time without simplification of the model. The trend in computing is toward parallel processing on shared-memory processors (SMP) instead of ever faster clock speeds. To get high-fidelity MBS models to run real-time it is necessary to adapt the solution to take advantage of multiple cores. In most real-time models the suspension compliance is modeled by the generation of look-up tables or elasto-kinematic maps, which replace the suspension geometry. The use of look-up tables makes it difficult to change parameters associated with bodies in the suspension. Inclusion of the suspension bodies and the compliance will require the use of implicit integration techniques in real-time solution to maintain a reasonable step size. This presents a challenge in real-time simulation since an implicit solver is not deterministic in that it takes a different amount of CPU time per time step.
This paper will document the changes necessary to convert an MBS solver into a deterministic solver that can execute much larger MBS models in real-time and then apply these techniques to an industrial automotive example.