Reconfiguration in Cyber-Physical Systems Sebastian Wtzoldt System - PowerPoint PPT Presentation

Reconfiguration in Cyber-Physical Systems Sebastian Wätzoldt System Analysis and Modeling Group Prof. Holger Giese

Motivation: Reconfiguration in Cyber-Physical Systems 2 ■ “ Cyber-Physical Systems (CPS) are integrations of computation with physical processes.” [Lee2008] ■ “[CPS] … embed software, which: □ Record physical data via sensors □ Affect physical processes using actuators □ Actively interact with physical and digital world □ Are connected with one another and in global networks □ Use globally available data and services” [acatech2011] Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Motivation: Reconfiguration in Cyber-Physical Systems 3 [acatech2011] Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Motivation: Reconfiguration in Cyber-Physical Systems 4 [IEEESpectrumNov2012] http://spectrum.ieee.org/green-tech/advanced-cars/all-aboard-the-robotic-road-train Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Motivation: Reconfiguration in Cyber-Physical Systems 5 [IEEESpectrumNov2012] http://spectrum.ieee.org/green-tech/advanced-cars/all-aboard-the-robotic-road-train Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Motivation: Reconfiguration in Cyber-Physical Systems 6 [IEEESpectrumNov2012] http://spectrum.ieee.org/green-tech/advanced-cars/all-aboard-the-robotic-road-train Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Motivation: Reconfiguration in Cyber-Physical Systems 7 Software adaptation is (1) the adaptation of a software system or (2) the processes and activities related to the adaptation of a software system . ■ Parameter adaptation vs. compositional adaptation □ “ Parameter adaptation modifies program variables that determine behavior. ” [McKinley2004] □ “[… ] compositional adaptation exchanges algorithmic or structural system components with others that improve a program’s fit to its current environment. ” [McKinley2004] ■ Static adaptation vs. dynamic adaptation [McKinley2004] ■ Internal adaptation vs. external adaptation [Salehie2009] [Musliner1999] [Schäfer2007] Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Points to Discuss Reconfiguration in Cyber-Physical Systems 8 (Adaptation) ■ (I) cover dynamic behavior via explicit modeled and coexisting feedback loops ■ (II) reduce complexity and enable interaction via abstraction using runtime models following the model driven engineering approach ■ (III) consider specific domains and nonfunctional properties ■ (IV) support concurrent and distributed interactions of subsystems Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Points to Discuss: Cyber-Physical Systems Laboratory 9 Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Feedback Loop 10 Adaptable Cyber-Physical System Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Feedback Loop 11 Adaptation Engine A nalyze P lan Knowledge M onitor E xecute Adaptable Cyber-Physical System Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Feedback Loop with Runtime Models ■ External adaptation approach 12 Adaptation Engine as proposed in [Salehie2009] ■ Feedback loop with four A nalyze P lan activities: □ Monitor-Analyse-Plan- Execute  Knowledge (MAPE-K) [IBMKephart2003] Knowledge ■ Runtime models as on-line representation of the running system [Vogel2011] □ Reflection models M onitor E xecute □ Evaluation models □ Change models Adaptable Cyber-Physical System □ Execution models Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Multiple Feedback Loops 13 Separation of loops: ■ Different concerns [Frey2012] □ Self-healing □ Self-optimizing □ Hard real-time adaptation vs. soft real-time A P A P adaptation ■ Local vs. global adaptation Self- Self- M E M E [Gueye2012] healing optimizing Adaptable Cyber-Physical System Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Multiple Feedback Loops 14 Problems: Separation of loops: ■ Inter-loop communication ■ Model exchange ■ Different concerns [Frey2012] ■ Concurrency coordination □ Self-healing ■ Contradicting model □ Self-optimizing manipulation □ Hard real-time adaptation vs. soft real-time A P A P adaptation ■ Local vs. global adaptation Self- Self- M E M E [Gueye2012] healing optimizing Adaptable Cyber-Physical System Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Multiple Layered Feedback Loops 15 … ■ Handle different problems on A P different level of abstraction e.g. inspired by adaptive control theory Self- M E [Kokar1999] optimizing ■ Hierarchical control architectures [IBMKephart2003] A P A P ■ Proposed reference architecture for self-managed software systems Hard Self- M E M E [Kramer2007] Real-Time healing Adaptation Adaptable Cyber-Physical System Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Distributed Multiple Layered Feedback Loops 16 … ■ Complex CPS with several independent, heterogeneous A P subsystems □ Autonomous robots in a Self- M E warehouse optimizing □ Distributed traffic management ■ Runtime models □ Basic communication concept □ Exchange of information Self- possible HRT HRT H. □ Learn, adapt on new situations via model driven techniques … Adaptable CPS Adaptable CPS Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Example: Cyber-Physical Systems Laboratory 17 Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Example 18 Adaptable Cyber-Physical System Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Example: Failure Detection and Repair 19 A P Failure Detection M E + Repair Adaptable Cyber-Physical System Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Example: Failure Detection and Repair  Monitor A P 20 Failure Knowledge Detection M E + Repair M onitor <<Sensor>> Localization <<Component>> Navigation Logic <<Sensor>> Obstacle Detection <<Actuator>> Wheel Adaptable Cyber-Physical System Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Example: Failure Detection and Repair  Monitor A P 21 Failure Knowledge Detection M E + Repair M onitor <<Sensor>> Localization <<Component>> Navigation Logic <<Sensor>> Obstacle Detection <<Actuator>> Wheel Adaptable Cyber-Physical System Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Example: Failure Detection and Repair  Analyze A P 22 Failure Knowledge Rules Detection M E + Repair A nalyze <<Sensor>> Localization <<Component>> Navigation Logic <<Sensor>> Obstacle Detection <<Actuator>> Wheel Adaptable Cyber-Physical System Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Example: Failure Detection and Repair  Analyze A P 23 Failure Knowledge Rules Detection M E + Repair A nalyze <<Sensor>> Localization <<Component>> Navigation Logic <<Sensor>> Obstacle Detection replace <<Actuator>> Wheel Adaptable Cyber-Physical System Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Example: Failure Detection and Repair  Plan A P 24 Failure Knowledge Detection M E + Repair P lan <<Sensor>> Localization <<Component>> Navigation Logic <<Sensor>> Obstacle Detection replace <<Actuator>> Wheel Adaptable Cyber-Physical System Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Example: Failure Detection and Repair  Plan A P 25 Failure Knowledge Detection M E + Component Repair repository <<Sensor>> P lan Localization <<Component>> Navigation Logic <<Sensor>> Obstacle Detection replace <<Actuator>> Wheel 1) Remove old component 2) Load backup component 3) Connect backup <<Sensor>> Adaptable Cyber-Physical System component BACKUP Obstacle Detection Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP

Example: Failure Detection and Repair  Execute A P 26 Failure Knowledge Detection M E + Repair <<Sensor>> E xecute Localization <<Component>> Navigation Logic <<Sensor>> Obstacle Detection replace <<Actuator>> Wheel 1) Remove old component 2) Load backup component 3) Connect backup <<Sensor>> component BACKUP Adaptable Cyber-Physical System Obstacle Detection Sebastian Wätzoldt | Reconfiguration in Cyber-Physical Systems | HPI Symposium @ SAP