Toolchain for a Mobile Robot Applied on the DLR Scout Rover

Abstract

This paper motivates and details the tools used to control a planetary exploration rover prototype and its simulation twin. The novelty is the coupling of the prototype and the simulation. The prototype and simulation of the rover use the same software tools and code for communication, signal processing, data storage and drive control algorithms. The presented toolchain is unprecedented for mobile robot to the best of the authors' knowledge, yet generic and versatile enough that it can easily be adapted to different kinds of robotic systems as the full paper will show. Advantages and drawbacks as well as comparison to other solutions are also key content of this text. DLR's Institute of System Dynamics and Control (SR) is developing a new mobile robot called \textquotedblScout\textquotedbl that will serve as the example of the toolchain. The Scout rover is a robust cave exploration rover, based on rimless wheels and biologically inspired locomotion mechanisms. It consists of segments interconnected by elastic \textquotedblvertebrae\textquotedbl in longitudinal direction. Some of the special features of this rover design can only reasonably be unfold thanks to the toolchain, e.g. the bionic phase shift control or simulative exploration of different rover and wheel configurations. The development of the Scout rover relies on modeling, simulation and optimization and follows the \textquotedblV\textquotedbl-shaped paradigm of system development. Elements of rapid control prototyping are implemented and commercial of the shelf sensors, actuators and computation units integrated. This has rapidly led to a robust mobile robot that is capable to cross obstacles that are larger than the radius of the wheels. The Scout rover development relies on the Modelica language for guiding and safeguarding virtual design decisions as well as for rapid control prototyping using the commercial Modelica modeling and simulation environment Dymola. Using this toolchain is a good compromise for satisfying the requirements in terms of rapid control prototyping. It has effective facilities for modeling the challenging physics for reliable Model-in-the-Loop simulations, at the same time the model of the controller can be directly used on the prototype's hardware. A well-thought-out model structure allows a transparent and seamless swapping between virtual model components and model components which drive physical drives and sensors. The actuators of the prototype are addressed using the EtherCAT protocol, communication with sensors and human operators goes through a light-weight in-house middleware and logged data is stored in the popular HDF5 format. Coordination of the actuators and reaction to user input in a state machine like manner is implemented with scripts in the Lua language. Steps for further work include integration of cameras and implementation of obstacle detection, additional control algorithms and more autonomy. Operations in the field of a space mission is envisioned around 2030. The presented toolchain is thought-out enough that this won't require major changes even when COTS parts need to be replaced by space qualified hardware. Likewise, the simulation of the rover will remain an important activity.