9/3/2008 Networked Embedded Systems School of Electrical Engineering Royal Institute of Technology Stockholm, Sweden Karl H. Johansson, Mikael Johansson, Carlo Fischione, Henrik Sandberg, Dimos Dimoragonas, Maben Rabi, Alexandre Seuret Piergiuseppe Di Marco, Erik Henriksson, Björn Johansson, Magnus Lindhe, Chithrupa Ramesh, Pablo Soldati, Pan Gun Park, Per Sahlholm, Lei Bao Networked Embedded Systems Embedded systems are getting networked Major implications on design, implementation and operation: + Cost efficient deployment and operation Increased system complexity Shared and dynamic network resources Requires new integrated tools from control, communication and computer sciences Control over wireless networks How control a plant when sensor, actuator and controller nodes are wireless network devices? Sensor and actuator network applications Industrial automation Home automation Transportation networks Home Network Wireless communication links Sensors Controllers Actuators Marine habitat mapping Surveillance Lindhe & Johansson, 2008 Environmental monitoring Plant FeedNetBack Economist 1
9/3/2008 A communication or a control problem? Approaches to control over wireless networks 1. Communication protocol suitable for control 2. Control application that compensates for communication imperfections 3. Integrated design of application and communication layers Control Application NET MAC PHY WirelessHART New wireless networking protocol designed for control applications Actuator Plant Sensor Wireless network Controller TDMA and CSMA time slots Typically 10 ms Periodic superframes of 100 time slots Networked Embedded Systems School of Electrical Engineering Royal Institute of Technology Stockholm, Sweden Karl H. Johansson, Mikael Johansson, Carlo Fischione, Henrik Sandberg, Dimos Dimoragonas, Maben Rabi, Alexandre Seuret Piergiuseppe Di Marco, Erik Henriksson, Björn Johansson, Magnus Lindhe, Chithrupa Ramesh, Pablo Soldati, Pan Gun Park, Per Sahlholm, Lei Bao 2
Main Competence LECS / ICT in ICES Axel Jantsch School of ICT, ICTKTH SoC Design Architecture Modeling Analysis, functional validation, performance analysis Techniques, tools, methodologies NoC FPGA ipack Center of Excellence: Sensor networks RF sensors, ids, communciation Innovative packaging Paper based electronics LECS/ICT: Component and Platform Provider Components and Platforms ITM Components: SoC, FPGA, Sensors, Actuators, Embedded SW, Architecture: Application specific EES CSC Design Methods and Tools: Modeling, Design, Synthesis, Verification, Components&Platforms: SoC, FPGA, MPSoC, Link Level Communication (Radio, wire, ), Sensors, Actuators ipack CMOS Packaging Post CMOS 1
Performance Analysis of Adaptive Systems Architectural Model Mode Adaptivity Just in Time Adaptivity 2
Just in Time Adaptivity with Prefetching Adaptive Streaming Streaming model is based on self-timed execution model, i.e. process executes only if there is enough space at the output buffer and there are enough tokens in the input buffer. Performance Analysis Framework Network Calculus to model data and processing rates Integer Linear Programming to find minimum buffer sizes Simulation of Synchronous MoC to identify a feasible periodic schedule Design cost: JIT and Prefetch 3
Summary Performance analysis for adaptive systems and reconfigurable architectures Framework with hybrid analysis Integer Linear Programming Token based simulation with ForSyDe Synchronous MoC 4
Computer Vision and Active Perception Laboratory (CVAP) Computer Vision: Stefan Carlsson Robotics: Danica Kragic KTH/VR Danica Kragic Lars Bretzner Oscar Linde Oscar Danielsson Per Rosengren Andrej Pronobis Babak Rasolzadeh VISCOS-SSFSSF Stefan Carlsson Josephine Sullivan VISUAL-KKS/SSF Stefan Carlsson Josephine Sullivan Martin Ericsson COSY-EU (Henrik Christensen) Patric Jensfelt Kristoffer Sjö Projects and People CAS-SSF Jan-Olof Eklundh Patric Jensfelt Danica Kragic PACO-PLUS-EU Jan-Olof Eklundh Danica Kragic Mårten Björkman Hedvig Kjellström Kai Hubner Javier Romero Gonzalez MOBVIS-EU Jan-Olof Eklundh Alireza Tavakoli FFL-SSF Danica Kragic Maria Ralph Carl Barck-Holst NEUROBOTICS-EU (Henrik Christensen) Christian Smith Mattias Bratt CogX-EU Patric Jensfelt Danica Kragic Adrian Bishop Gareth Loyd Yasemin Berkiroglu Alper Aydemir GRASP-EU Danica Kragic Patric Jensfelt Dan Song Jeannette Bohg Niklas Bergström COGNIRON-EU (Henrik Christensen) Elin Anna Topp Camera Tactile Sensors Laser Sonar Gyro Accel. What CVAP does Sensor processing Recognition Scene analysis Localization Grasping Human-Robot collaboration Recognition: Finding specific objects Learning by demonstration 1
Recognition: Finding your position GPS coordinates Tracking and labelling collective human motion Image database (e.g. panoramas) Start up company CogEye AB User s image User s location Position estimation Human Robot Collaborative Systems Surgical systems capable of modeling and following the progress of a surgical procedure, and able to use robotic devices and sensors extending human performance cooperatively with the surgeon in order to improve the quality of surgical procedures and to enable interventions that would be otherwise impossible. 2
Grasping and Manipulation Hardware Things Objects Attention points 2D Attributes 3D Segmentation 3D Attributes Grouping 2D/3D Attributes Course: Robotics and Autonomous Systems Build and program a robot 3
Spin offs 3 startup companies in the last 4 years: CogEye: wide screen video panoramas for TV production InMoDo: image applications on mobile phones Intelligent Machines: technology and solution provider for robotic applications CogEye AB cont d KKS/SSF: 3D visualization of real world football games in collaboration with TRACAB (Solna) Technology and solutions provider to the emerging service robot industry 4
ICES 20080903 Architecting dynamically configurable embedded systems Embedded Control Systems Research group Mechatronics, Machine Design, ITM Martin Törngren, DeJiu Chen, Lei Feng, Tahir Naseer, Magnus Persson and Javier Garcia www.dyscas.org 0 Architecting challenges Expanding systems! More functions, sensors, actuators, embedded computers and internal networks External connections (internet, cooperative systems) Component reuse, product lines Evolving systems! Changing hardware and software platforms Software upgrades Later systems integration, adding devices Integrating and removing functions/devices in use Internal reconfigurations to deal with varying, partly unknown, and changing systems Additional complexity calls for systematic approaches and new paradigms to handle run-time reconfiguration. ICES kick-off - 2008-09-03, KTH Martin Törngren, Mechatronics 1 1
ICES 20080903 Needs and challenges for run-time reconfigurations Traditional embedded systems are statically configured Run-time dynamics/changes are caused by Varying load, execution times and interference Down and upgrading of products in use o Software, Devices and Functions Internal reconfigurations Challenges Ensuring efficient and predictable run-time reconfiguration Validation and context dependencies Security ICES kick-off - 2008-09-03, KTH Martin Törngren, Mechatronics 2 Dynamically configurable systems Integrating/ updating SW functionality via hotspots Discovery and integration of new devices attached Closed Reconfiguration, for performance or reliability purposes ICES kick-off - 2008-09-03, KTH Martin Törngren, Mechatronics 3 2
ICES 20080903 The DySCAS Architecture 1: Context monitoring & event detection 2: Reasoning and decision 2a: Dynamic Configuration Control 2b: Dependability & QoS 3: Actuation ti and synchronization 4: Transparency 4 1 2a 2b 3 ICES kick-off - 2008-09-03, KTH Martin Törngren, Mechatronics 4 Model based engineering for dynamically configurable systems Requirements Hazard, FTA and FMEA Behavior design, e.g. Matlab/Simulink and evaluation Formal verification S a f e t y Vehicle Level Analysis Level Implementation Level Operational Level Additional views Configuration, Dynamic envelope Database ICES kick-off - 2008-09-03, KTH Martin Törngren, Mechatronics 5 3