Tech Briefs

Integrated simulation capabilities that are high-fidelity, fast, and have scalable architecture are essential to support autonomous vehicle design and performance assessment.

Integrated simulation capabilities that are high-fidelity, fast, and have scalable architecture are essential to support autonomous vehicle design and performance assessment for the Army’s growing use of unmanned ground vehicles (UGVs). A mobility simulation of an autonomous vehicle in an off-road scenario was developed using integrated sensor, controller, and multibody dynamics models.

Figure 1. The HMMWV Simulation Model built in ROAMS.

The JPL Rover Analysis, Modeling and Simulation (ROAMS) ground vehicle simulation framework is based on the JPL Darts/Dshell simulation architecture. ROAMS and the underlying architecture have been successfully demonstrated at JPL in several space scenarios where a high degree of mission complexity, real-time performance, and extensive sensor/actuator/ control integration were necessary. ROAMS is unique in its integrated approach to straddling the multifunction, high-fidelity dynamics, sensors, environment, control, and autonomy models that are key attributes of future Army UGVs.

Figure 2. The vehicle during the Lane Change Maneuver. Note the roll of the vehicle as it changes lanes.

This work applies the ROAMS modeling approach to address the fidelity and speed bottlenecks for the Army’s need to evaluate and test autonomous ground vehicles. This scalable architecture allows the adaptation and tuning of simulation fidelity across a very broad range (e.g. rigid/flex-body dynamics, sensor fidelity, dynamics/kinematics modes) needed for the multi-layered testing of complex autonomy behaviors.

This project developed and demonstrated an integrated simulation capability consisting of real-time, high-fidelity dynamics with control, sensors, and environment models in the loop for a representative autonomous vehicle. The simulator’s architecture will allow the seamless selection of different fidelity levels and model parameters across the full modeling suite, and more importantly, provide analysts with a modular way to swap component models for varying vehicle/control/sensor behavior.

The HMMWV suspension system is a variant of the common double wishbone suspension. These suspension systems have a large number of distinct bodies that are contained within a single kinematic closed loop. As a result, these suspensions have a large number of internal degrees of freedom, but due to the constraints imposed by them to a frame or chassis, they only have a single independent degree of freedom. For the HMMWV suspension modeling, three algorithmic techniques were tested and benchmarked to solve the multibody dynamics of the suspension system. While the HMMWV suspension model is the same, the difference among the three techniques is the number of constraints that are needed, which directly affects their resultant computational speed.

The HMMWV vehicle model built in ROAMS has essentially 15 degrees of freedom (DOF) after taking into account all the constraints on the system (Figure 1). Vehicle simulations were run in two main environments. The first was an urban environment; the second was an off-road environment. In both cases, a graphical representation of the environment with a variety of tools was created. A digital elevation map was extracted from it for use in the vehicle wheel-terrain contact simulation.

The urban environment consisted of a 3D mesh model of a city. The vehicle offroad environment consisted of bumpy terrain (as opposed to the paved flat urban environment). To create the bumpy offroad environments, a digital elevation map was created and imported into the simulation. LIDAR was simulated using the simulator’s graphics system.

Three distinct scenarios were performed.

  • Scenario 1: Urban driving with navigation and obstacle avoidance. An urban terrain was used and a goal point a few hundred meters away was defined. The vehicle was left to drive towards the goal. The navigation system detected the raised edges of the road and avoided them.
  • Scenario 2: Urban driving - lane change maneuver. The vehicle was driven in the urban environment, and an open-loop lane change maneuver was performed at speed to demonstrate the realistic nature of the vehicle dynamics (Figure 2).
  • Scenario 3: Off-road driving teleoperation. The vehicle was driven at various speeds on off-road terrain with a few trees to demonstrate vehicle behavior.

The ROAMS HMMWV simulation successfully demonstrates that high-fidelity multibody dynamics, terrain models, sensors, actuators, control, and navigation in urban and off-road scenarios can be modeled and run at speeds that are useful for vehicle analysis and design purposes.

This work was done by Jonathan Cameron, Steven Myint, Calvin Kuo, Abhi Jain, and Hävard Grip of Jet Propulsion Laboratory, California Institute of Technology; Paramsothy Jayakumar, of the U.S. Army TARDEC; and Jim Overholt of the U.S. Air Force Research Laboratory. ARL-0162