Simulation

Simulation environments are representing an important tool for the development of autonomous robot controls. In the following, various simulation tools and their application in research are presented.

SimVis3D

SimVis3D is a 3D simulation software, which is developed for the robot framework Finroc. It is based on Open-Inventor and the Newton physics engine and provides simulated sensors such as cameras, laser scanners, infrared, ultrasound or PMD sensors. A simulation scene integrates arbitrary 3D models, which are imported via an XML file and transferred to a scene graph. At runtime, a user can interact with the scene and change it via C ++ interfaces at runtime. The software was used to simulate a wide variety of robots, such as the mobile robots RAVON, MARVIN, ARTOS, the excavator THOR, or the running machine BIPED.

With SimVis3D the framework for the autonomous testing TAURUS was developed. It enables the specification of complex test cases. TAURUS tests consist of a scenario description that contains the navigation commands and a "mission" the robot has to perform. Various parameters of the simulated robot and its environment are recorded and used for the evaluation. If an invariant (e.g., timeout, collision, ...) is detected, this information is provided in a report after each run. TAURUS itself consists of a behavior network, which is connected between the running program and the user interface. It automatically controls the various test cases, sets parameters and configurations.

Ogre-PhysX

The Ogre-PhysX simulation was developed as a successor to the SimVis3D software. It is based on the graphics framework OGRE3D and NVIDIA PhysX and offers advanced performance and improved graphics. Complex search & rescue scenarios were simulated, in which complete houses with thousands of objects were dynamically simulated to carry out autonomous rescue and reconnaissance missions. Danger sources such as fire or unstable environmental segments were mapped with high performance, so that the robot controller could be optimized to adequately handle such situations. In addition to this scenario, equally large open worlds were implemented in which autonomous rides over long distances were made possible and farming processes were simulated.

The simulation was used for the control processing of the robots Gator, SUGV, and LUGV.

V-REP

The Virtual Robot Experimentation Platform (V-REP) is a simulation engine of Coppelia Robotics. It supports, among others, C / C++, Python, Lua, Octave, Urbi and Matlab. To simulate physics, various engines are available, such as Newton, Bullet, ODE or Vortex. V-REP interacts with the Finroc Framework by the use of a C++ interface. Finroc provides a V-REP plugin for this, but can also communicate via the Remote API.

Among other things, V-REP was used for the simulation of the Gator, Unimog, Backhoe and CARL. In addition to sensors such as 2D and 3D lasers, GNSS and IMU, characteristic sensor disturbances and effects were also implemented. Thus, for the determination of GNSS signals, real satellite data can be read out from Finroc and  noise by shadowing effects can be realistically calculated.

Unreal Engine 4

The Unreal Engine 4 is a professional game engine developed by Epic Games. The platform offers photorealistic rendering, VR connectivity, open world design, blueprint scripting and full C++ source access. Projects like NVIDIA FleX complement the existing PhysX engine with particle simulations. A marketplace provides access to a number of projects and content.

Finroc is realized as a Unreal project which starts the Finroc runtime environment and ports from the engine. It is possible to control and interact with robots in a complex, open, photorealistic world. The detailed simulation of sensor technology includes for example, typical camera-fades, -noise, or -blurring which have a wide influence on the controller in real systems.

Unreal is currently being used in road construction scenarios with the road rollers BW154 and BW147 for the AMMCOA project, where a few kilometers of the federal highway B10 are reconstructed and simulated. Likewise, Gator, Unimog, Backhoe and the environment detection of the mobile cranes of the SafeguARd project are tested virtually.

Relevant Publications

@inproceedings{Braun07,
Author = {Tim Braun and Jens Wettach and Karsten Berns},
Title = {A Customizable, Multi-Host Simulation and Visualization Framework for Robot Applications},
Booktitle = {Proceedings of the 13th International Conference on Advanced Robotics (ICAR07)},
Year = {2007},
Pages = {1105-1110},
Address = {Jeju, Korea},
Month = {August 21-24},
}