Enhanced Simulation Techniques for the Automatic Evaluation of Vehicle Designs

Downs, D., Comer, T., Archibald, C., Bethel, C. L., Goodin, C., & Durst, P. (2016). Enhanced Simulation Techniques for the Automatic Evaluation of Vehicle Designs. SAE 2016 World Congress and Exhibition in the Modeling and Simulation of Military Ground Vehicles (IDM201) track. Detroit, MI.

The ability to quickly and automatically evaluate vehicle designs is a critical tool in an efficient vehicle design process. This paper presents techniques for vehicle parameter estimation using automatic intelligent simulations. These techniques enable the efficient and automatic evaluation of many important aspects of vehicle designs. The effectiveness of this approach is demonstrated by using vehicle tests that are commonly performed on military ground vehicles. Our simulation techniques are able to determine the relevant vehicle performance characteristics in a much more efficient manner than could be done previously. This is done automatically, once the user has specified the type of test to be performed. A terrain sample is automatically generated and the vehicle’s behavior on each terrain instance is evaluated until the specified test conditions are met. A search algorithm determines which terrain to generate at each iteration, and the effectiveness of two different search algorithms are investigated for this purpose. Simulation also allows for efficient measurement of new performance characteristics that are impractical to measure in the real world. The described techniques have been implemented as part of a software plugin for the ANVEL simulation environment. This plugin gives the ability to easily obtain vehicle performance metrics that reflect the design's performance over a wide range of possible physical environments. Using this framework users can easily implement whichever new tests they are interested in performing. This framework will be made freely available so that other designers can obtain vehicle evaluations in a timely manner and with minimal user interaction.