Computer Assisted Engineering for Robotics and Autonomous Systems - PowerPoint PPT Presentation

Computer Assisted Engineering for Robotics and Autonomous Systems Development and Adoption of Model-Based Tools in Robotics Christian Schlegel Alex Lotz, Matthias Lutz, Dennis Stampfer Computer Science Department University of Applied Sciences Ulm, Germany http://www.servicerobotik-ulm.de/ https://www.youtube.com/user/RoboticsAtHsUlm Christian Schlegel, Schloss Dagstuhl, 13.02.2017 1

iserveU hospital logistics iserveU Final Report Christian Schlegel, Schloss Dagstuhl, 13.02.2017 ECHORD++ MARS Experiment http://www.fendt.com/int/11649.asp 2 Intralogistics OpenRobotino KMU Innovativ

https://youtu.be/KH7FJnNO3z4 Christian Schlegel, Schloss Dagstuhl, 13.02.2017 3

Cooperative Robot Butler Scenario Christian Schlegel, Schloss Dagstuhl, 13.02.2017 4

https://youtu.be/DjjNUPpj36E Christian Schlegel, Schloss Dagstuhl, 13.02.2017 5

Development and Adoption of Model-Based Tools in Robotics Paper describes how Boeing built the 777 with thousands of workers, with many companies around the world, doing everything by software (from design over manufacturing to checking the software) IEEE Spectrum, October 1995 Resulted in tremendous cut in costs and cut in time to manufacturing Christian Schlegel, Schloss Dagstuhl, 13.02.2017 6

Development and Adoption of Model-Based Tools in Robotics The Beauty in Systems Engineering… Where we are … Existence Proof: • from CAD schematics to chips How biology builds complex systems … • Boeing 777 (1986-1994) / Boeing 787 (2004-2009) • Cyber-Physical Systems (CPS) (2008…) • Social networks mushroomed over the Web (based on John S. Baras / 2003 White Symposium) • Economic networks mushroomed over the Web • Renewable energy, smart grids • … organism organ tissue cell organ system information + software + physical components from DNA „ programs “ + humans to „ devices “ / living organisms Overall challenge: we need to provide technical solutions that • provide great performance • are safe, secure, robust, resilient, predictable, conformant to legal and ethical norms, etc. • are affordable, economically justifiable, manageable etc. Christian Schlegel, Schloss Dagstuhl, 13.02.2017 7

Computer Assisted Engineering for Robotics and Autonomous Systems Development and Adoption of Model-Based Tools in Robotics • computer aided engineering • computer assisted engineering • computer assisted software engineering Christian Schlegel, Schloss Dagstuhl, 13.02.2017 8

Development and Adoption of Model-Based Tools in Robotics Topic Group Software Engineering, System Integration, System Engineering Christian Schlegel, Schloss Dagstuhl, 13.02.2017 9

Development and Adoption of Model-Based Tools in Robotics Modeling languages that are not executable, or where the execution semantics is vague or undefined are not much better than TUML: • interested in executable models only (based on Edward Lee,Berkeley, http://models2010.ifi.uio.no/material/Heterogeneous_Models.pdf) (not in descriptive models, declarative models fine as long as they come with solvers) • focus on concurrent components that communicate via ports (as one might describe in SysML, AADL, etc., though I will be more specific than these) We can do better! We need do better! The Truly Unified Modeling Language TUML Examples of TUML languages: • notice how nicely formal the language is! • tools already exist • with the mere addition of a TUML profile, every existing UML notation is a special case! Christian Schlegel, Schloss Dagstuhl, 13.02.2017 10

Development and Adoption of Model-Based Tools in Robotics Example ROS: Purely source code-driven, implementation-driven ecosystem http://design.ros2.org/articles/why_ros2.html Christian Schlegel, Schloss Dagstuhl, 13.02.2017 11

Development and Adoption of Model-Based Tools in Robotics • AUTOSAR : timing constraints meant to describe timing chains for the main purpose of worst-case timing analysis: • not for specifying timings • not related to QoS (no link to implications of modified cycle times etc.) • not to support deployment source: Marco di Natale Fine-grained • … functional graphs • MARTE profile mapped onto (OMG standard for real-time and embedded systems): complex distributed • too large architecture • difficult to use • union of needs/requirements from too many sources • OMG MDA (model-driven architecture): • too linear workflow PIM – > PSM – > PSI • Implementations abstract away relevant properties like QoS, resource awareness, etc. • … Christian Schlegel, Schloss Dagstuhl, 13.02.2017 12

Development and Adoption of Model-Based Tools in Robotics • scientific modeling • models in engineering • building a model requires abstraction • assumptions are used in modeling in order to specify the domain • structural model of the system: what the system consists of? => robot base, manipulator, laser ranger, SLAM, object recognition, HRI, … • behavioral model of the system: what the system does? => deliver coffee, grasp cup, operate coffee machine, … All can be block diagrams with hierarchy. The difficult part is the mapping between behavior and structure. • a system is a set of interacting or independent entities, real or abstract, forming an integrated whole • composability is the ability to combine and recombine as-is building blocks into different systems for different purposes. It requires that properties of sub- systems are invariant („remain satisfied“) under composition. • splittability is the „inverse“ relationship of composability. • compositionality requires that the behavior of a system is predictable from its sub-systems and that of the composition „glue“. • system composition (activity): the activity of putting together a set of existing building blocks to match system needs with a focus on flexible (re-)combination. • system integration (activity): the activity that requires effort to combine components, requiring modifications or additional actions to make them work with others. Christian Schlegel, Schloss Dagstuhl, 13.02.2017 13

Development and Adoption of Model-Based Tools in Robotics Can we think of complex robotic systems before we start constructing them? • there will be no single architecture („ the “ reference architecture) as „ architecture “ is for something • architectures are linked to properties like resilience, robustness , … • do not abstract away relevant aspects • resources • extra-functional properties ( how to execute a functionality) • do not ignore relevant means to deal with complexity • separation of concerns • composability How to build models from all the heterogeneous sciences for different parts and aspects of a robotic system?  be able to put together the models / link the models quickly and correctly  be at least as detailed as needed for a certain level of confidence into the properties of the outcome (by simulation, by testing, by reasoning , …) You cannot go through all combinations of all parameters in order to know about the properties of your system. Your need to be able to answer „ what if “ questions with design exploration tools which give the answers quickly and are user-friendly (trade-off analysis, multi-criteria-optimization, constraint-based reasoning all composed) Christian Schlegel, Schloss Dagstuhl, 13.02.2017 14

Development and Adoption of Model-Based Tools in Robotics • hook heterogeneous models • cannot be done easily as long as you do not adopt to a notion of Meta-Models • Meta-models allow to transform in a consistent way between models including constraints, tolerances etc. (John S. Baras / 2016 ETFA / https://youtu.be/oXAZEIZc5Zs?t=1354 ) Christian Schlegel, Schloss Dagstuhl, 13.02.2017 15

Development and Adoption of Model-Based Tools in Robotics horizontal / composable vertical / composable Christian Schlegel, Schloss Dagstuhl, 13.02.2017 16

Development and Adoption of Model-Based Tools in Robotics Christian Schlegel, Schloss Dagstuhl, 13.02.2017 17

Development and Adoption of Model-Based Tools in Robotics computation communication achieve separation of roles coordination support composition configuration Which patterns and structures form the Sweet Spot between Freedom of Choice and Freedom from Choice ? Guidance for separation of concerns by superordinate objectives like the need for separation of roles and the need for composition Support as much freedom as possible while still ensuring composability despite separation of roles User Christian Schlegel, Schloss Dagstuhl, 13.02.2017 18