|Title||Real-Time Simulation and Visualization Architecture with Field Programmable Gate Array (FPGA) Simulator|
|Publication Type||Conference Papers|
|Authors||M. Karkee, M. Monga, B. Steward, J. Zambreno and A. Kelkar|
|Conference Name||Proceedings of the ASME World Conference on Innovative Virtual Reality (WINVR)|
Real-time simulation of dynamic vehicle system models is essential to facilitate advances in operator and hardware in the loop simulation and virtual prototyping. Real-time virtual reality-based simulation enables users to visualize and perceive the effect of their actions during the simulation. As model complexity is increased to improve the model fidelity, the computational requirements will also increase, thus increasing the challenge to meet real-time constraints. A distributed simulator architecture was developed for off-road vehicle dynamic models and 3D graphics visualization to distribute the overall computational load across multiple computational platforms. This architecture consisted of three major components: a dynamic model simulator, a virtual reality simulator, and an interface to controller and input hardware devices. The dynamic model simulator component was developed using Matlab/Simulink Real Time Workshop on a PC and also using Field Programmable Gate Arrays (FPGA), which offered a highly parallel hardware platform. The simulator architecture reduced the computational load to an individual platform and increased the real-time simulation capability with complex off-road vehicle system models and controllers. The architecture was used to develop, simulate and visualize a tractor and towed implement steering dynamics model. The model also included a steering valve sub-system which contained very high frequency hydraulic dynamics and required 10 μs integration time step for numerical stability. The real-time simulation goal was not achievable for the model with this level of complexity when the PC-based simulator was used. However, the FPGA-based simulator achieved a real-time goal taking only 2 μs to complete one integration time step.