Talks

The Robot Operating System (ROS) is the largest open-source software project in robotics. This talk gives an introduction to ROS, ROS 2, and the unique history of Bosch's involvement with the ROS project and the open-source community, which began in 2010 with the PR2 Beta Program. It highlights typical patterns for the adoption of ROS in industry and describes how Bosch Rexroth integrates ROS 2 into the ctrlX OS and the open ctrlX AUTOMATION ecosystem. Additionally, it provides insights into the daily use of ROS in the development of software components for mobile robotics within Bosch Rexroth’s Robotics Kit (ROKIT).

With regard to verification and validation (V&V), the software engineering for autonomous robots is particularly challenging: the close integration with sensors, actuators, and mechanics, deployment and operation in potentially unknown environments, and the requirement to operate over long periods of time without human intervention. In this talk, typical approaches and processes in software engineering for autonomous robotic systems are explained, the particular challenges are elaborated, and an overview to modern methods for V&V in this domain is presented. Finally, current research questions are discussed and the methods and tools of the EU-funded CONVINCE project are introduced, which will be presented in detail later in the day.

Any open-source software must have a license. ROS packages declare this license in their package.xml. It is a considerable risk to the usage of ROS in commercial products if this information is insufficient or incorrect. For this reason, the package.xml format (REP-149) has recently been slightly extended and made more precise. We discuss these changes and introduce the ros_licensing_toolkit. It scans the source code of ROS packages and checks if the discovered licenses are declared correctly. Additionally, it supports the generation of copyright files according to the Machine-readable Debian Copyright file standard, which shall be integrated in the Build Farm soon.

The Bosch Group has a uniquely diverse set of robotic products and applications, from consumer, to professional service, to industrial robotics. The research topics in robotics being studied at the various Bosch Research locations around the globe are even broader. In the first part, this talk gives an overview to these topics including SLAM, human motion prediction, and MPC based motion planning for mobile robots, learning and planning for complex assembly tasks with dexterous joining processes, scene segmentation and learning for robot handling in warehouses, novel mechatronic elements for robotic insertion processes in mass production.

The second part covers research on robot software platforms and tools, with many contributions to the open source Robot Operating System. This includes for example micro ROS, execution management, and very new research on methods for model checking of robot deliberation with Behavior Trees. Finally, in the third part, a spotlight on the research on multi agent path finding for AMRs in intralogistics but also for contactless planar systems such as ctrlX FLOW6D is given.

Robotic assembly of advanced component parts (e.g., for automotive or automation industry) poses many challenges in dexterous manipulation. The increased need for flexibility raises further challenges in efficient programming of robots on the shop floor - aiming at optimal resource usage and cycle times. We’ll discuss these challenges by concrete examples and approaches from an industry's point of view and outline a concept for a holistic approach for programming complex assembly tasks by multi-domain optimization from multimodal specifications.

As one of the 11 participants of the PR2 Beta Program, Bosch has more than 12 years of history with ROS. Over this time, ROS has been adopted by numerous R&D projects, from early research to product development. At the same time, Bosch made numerous contributions, has shaped the development of ROS 2. This talk will give an overview to the different use-case and application domains for ROS at Bosch, discuss motivation and hurdles with the adoption of ROS, elaborate the collaboration with the global ROS community, and give an overview to selected technical and organizational setups.

Micro-ROS puts ROS 2 on microcontrollers under the assumption of a POSIX-compliant operating system. Special measures are required to put ROS 2 on automotive microcontrollers running AUTOSAR Classic: (1.) The ROS execution model must be mapped to the fixed periodic preemptive scheduler. (2.) The ROS Client Support Library has to be qualified for static memory allocation. (3.) The middleware must be extended for the prevalent CAN protocol. We present the necessary extensions and changes by the example of an off-road vehicle control unit running a 1kHz control application. Furthermore, we propose a performance model of micro-ROS for design space exploration.

High-level deliberation in a robotic system has to simultaneously handle three different but closely interwoven aspects: intended tasks, contingencies, and system-level errors. To support developers in handling this complexity, we developed the System Modes package. It provides a modeling concept and mechanisms for scalable runtime management and reconfiguration of large ROS-based systems. The System Modes package leverages the ROS 2 node lifecycle and parameters to allow specifying operational states and modes over multiple ROS nodes hierarchically. It provides services for monitoring and runtime management of the so modeled system hierarchy in a self-similar approach with the standard lifecycle services.

In this panel moderated by Tom Moore (Locus Robotics), I discussed lessons learned on the migration to ROS 2 with Brian Wilcox (iRobot) and Stephanie Eng (Picknik.ai).

The Functional Mock-Up Interface (FMI) is a standard (http://fmi-standard.org/) for exchange and co-simulation of dynamic models as functional mock-up units (FMU). FMUs can be exported from a number of modeling and simulation tools, including Dymola, MATLAB/Simulink, OpenModelica, SimulationX, and Wolfram System Modeler. In this talk, the new fmi_adapter package (https://wiki.ros.org/fmi_adapter) is presented, which allows to run FMUs in or as ROS nodes.

Robotic systems typically comprise different applications with varying real-time and predictability requirements – even within ROS nodes. Therefore, we propose an enhanced Executor concept for ROS 2.0 allowing to schedule callbacks on a fine granularity based on annotations with real-time profiles.