This dissertation addresses the development of an effective real-time speed selection methodology for autonomous ground vehicles operating at high-speeds on rough outdoor terrains for the purpose of traveling as fast as possible without vehicle or payload damage.

A novel real-time capable simplified vehicle model is presented for simulating and predicting vehicle dynamics. For the purpose of performing simulation validation, a developed V-REP multibody simulation model of the experimental ATV used in this dissertation is combined with the MF-SWIFT advanced tire model, which is called the V-REP ATV model. These models are compared to experimental results.

A real-time capable tire model for computing longitudinal, lateral, and vertical compressive tire forces in real-time is also developed, called the Centroid Method. Compared to the experimentally tested advanced tire model MF-SWIFT, this Centroid method performed well with good performance for the longitudinal and vertical tires forces up to 8-21Hz.

For the purpose of parameterizing tire vertical stiffness in the absence of load testing equipment, the novel optimization method for determining the shape of a tire encountering terrain was developed called Tire Air Volume Maximization using Penalty Constraint Optimization (TAVM-PCO). This was combined with the existing Energy method for estimating tire forces from tire shape. This approach showed good performance for estimating tire vertical stiffness, and requires only tire geometry and tire pressure to be measured.

Using the developed vehicle and tire models, a speed selection methodology is developed using predictions of upcoming change in vertical kinetic energy of the vehicle, as well as work done by horizontal tire forces on the vehicle. By determining allowable limits for work done and changes in kinetic energy, speed is selected to maintain safe vehicle behavior over upcoming terrain. This speed selection methodology is both tested in simulation against the V-REP ATV model, and compared to the experimental ATV data.