UNIVERSITY OF IOWA

Virtual testing of autonomous mobility systems requires accurate and computationally efficient predictions of tire–soil interactions across diverse terrains and soil conditions. This talk introduces a novel grid-based, data-driven transient tire–soil contact model that combines the physical structure and interpretability of classical terramechanics with neural networks to capture complex, transient soil behavior. The proposed approach aims to bridge the gap between computational efficiency and high-fidelity predictions, with its accuracy and performance evaluated against field test data.