Skip Menu

Return to Skip Menu

Main Content


Development of Virtual Proving Ground Software
Chrysler 2013-2014 Advisor:Dr. John Ferris

This research builds on the long-standing successful relationship between Chrysler and Virginia Tech’s Vehicle Terrain Performance Laboratory (VTPL).  Previous work includes high-fidelity, high-resolution measurements of Chrysler’s Chelsea Proving Grounds (CPG) in 2007 and 2010.  Three new software packages were delivered at the conclusion of the 2011 and 2012 research projects: XYZTools (anomaly correction software to parse, view, and edit large datasets), TerrainSim (terrain surface characterization, modeling, analysis, and synthesis software paid for by the US Army), and most recently CloudSurfer.

CloudSurfer incorporates state‐of-the‐art algorithms and computationally efficient parallel processing to develop a ‘curved regular grid’ of terrain surface data, delivered in a number of useful formats that can be read directly into ADAMS (CRG, RGR, RDF, and the FIAT specific), and runs on either Windows or LINUX platforms. CloudSurfer generates this gridded data based on a center path along the road that is either generated automatically or prescribed by the user.  Additional alternate paths can also be automatically generated (so that path variation can be accounted for in vehicle simulations).

The current research project builds on the existing CloudSurfer Graphical User Interfaces (GUIs), efficient computing methods, and multiple platforms and formats.  The specific deliverables for the 2013 proposal are:

1.       A Chrysler-Specific Roughness Measure similar to the International Roughness Index (IRI), but developed specifically for different classes of Chrysler vehicles.  This would give an overall roughness level of a given road.  Specifically, a set of rainflow counting methods typically used in fatigue prediction could be employed.

2.       An adjustable roughness level for gravel roads at CPG.  That is, synthetic gravel roads would be generated that have a user prescribed level of roughness (perhaps on a scale from 0-flat to 10-extremely rough)

3.       Characterizing events.  This follows on the work completed on event detection.  A set of parameters (length, width, depth,…) will be used to characterize each event so that the number of each type of event can be counted.  This could then be used as an aid in developing Test Schedules in addition to the overall roughness level of a given road.

Developing terrain meshes for use in multi-body dynamic (MBD) and finite element (FE) models from either a specific real-world location or from a statistically representative virtual location.(VTPL Industrial Affiliates Program, current member: John Deere). Advisor:Dr John Ferris

In order to do virtual testing of agricultural machines, engineers need to be able to re-create customer usage in simulations.  The shape, slope, and roughness of the terrain are important model inputs.  The Vehicle Terrain Performance Laboratory has developed two programs (CloudSurfer and TerrainSim) to transform measured data to useful information as outlined in the flowchart below.



This research focuses on several enhancements to this software, specifically

  • Removing instrumentation drift from point cloud data.
  • Joining data from geographically adjacent locations after being adjusted for instrumentation drift.
  • Developing new data acquisition procedures that work in concert with the data processing techniques
    (This may include using calibration points on the ground to provide reference points for drift removal).
  • Removing vegetation from point cloud data.
  • Enhancing TerrainSim to accept a wider array of  file formats
  • Performing statistical analysis on the data that are relevant to non-directional (isotropic) terrain.  
  • Detecting and categorizing events in the terrain, such as: border crossings, levees, and ditches
  • Developing a method to generate synthetic,non-directional terrain based on a group of similar terrains.
  • Developing an Off-Road Roughness Measure appropriate for agricultural equipment

Nonlinear Predictor-Based Output Feedback Control of Uncertain Nonlinear Systems with applications to biomechanical systems Graduate Student:Chuong Nguyen  Advisor:Dr. Alexander Leonessa

 This research aims to develop a model-free controller for unknown systems. Common control techniques require good knowledge of the system as well as measurable states. In particular, when a system is observable, the state variables can be reconstructed by using observer techniques. However, most of such techniques also require that the system’s parameters are well defined in order to be applied.  In practice, full knowledge of system’s dynamics and complete measurable states is hardly achievable. For this reason, it is desirable to develop a controller that does not rely on the system knowledge. Well-known techniques such as PID, neural network and fuzzy control are used to deal with these systems; however, they must undergo a time-consuming tuning and training process. Our goal is to develop a different approach to control unknown systems with limited output measurements. The novel idea is that instead of trying to reconstruct the system model and its state, the controller relies on a predictor, which merely explores the output and input history to predict the immediate next output. Using Lyapunov stability theory, the predictor is designed so that its output will converge to the system output for any admissible choice of control input. Next, since the predictor is a completely known virtual system, using any available full-state feedback controller to drive the predictor output to a desired trajectory will simultaneously lead to the convergence of the actual system output to the same desired trajectory. This research has a large range of applications, and the current focus is to develop and test the controller for neuroprosthetic devices, such as using Functional Electrical Stimulation to assist limb motion for stroke and Parkinson patients.