Date post: | 10-Mar-2018 |
Category: |
Documents |
Upload: | hoanghuong |
View: | 213 times |
Download: | 1 times |
Information Modeling
Fast forward
Five posters
− Four introduced here
− One on Haptics project introduced by Jonas Unger
Three additional papers
Look-ahead Control
It is shown that an energy formulation allows
− simple integration routines
− interpolation on coarse grids
A residual cost is proposed
− derived from engine and driveline characteristics
Proper physical model of gear shift
Hellström, E., Åslund, J., and Nielsen, L. Design of an efficient algorithm for on-board look-ahead control. Solicited for Control Engineering Practice.
Look-ahead Control
In many modern engines the fuel map is non-linear due to the different control strategies that are used today
Different formulations of the fuel minimization problem are studied in this paper and consequences of the non-linearities are analyzed
Ivarsson , M., Åslund, J., and Nielsen, L. Optimal Speed on Small Gradients –Consequences of a Non-Linear Fuel MapIFAC World
Regression on ManifoldsFrom Shells to Temperature
14 M
ay-0
9
Chemical compositionmeasurements
Reconstructed water temperature
Shells from an archaeological site
• Henrik Ohlsson and Lennart Ljung. Gray-Box Identification for High-Dimensional Manifold Constrained Regression. In the15th IFAC Symposium on System Identification, SYSID 2009.• M. Bauwens , H. Ohlsson, K. Barbe, V. Beelaerts, F. Dehairs and J. Schoukens. On Climate Reconstruction using Bivalve Shells: Three Methods to interpret the Chemical Signature of a Shell, In
the7th IFAC Symposium on Modelling and Control in Biomedical Systems. 2009.• M. Bauwens, H.Ohlsson, K. Barbe, V. Beelaerts, J. Schoukens and F. Dehairs. Three Ways to do Temperature Reconstruction Based on Bivalve-Proxy Information. In the 28th Benelux Meeting on
Systems and Control, 2009.• Henrik Ohlsson, Jacob Roll and Lennart Ljung. Manifold-Constrained Regressors in System Identification. In Proceedings of the 47th IEEE Conference on Decision and Control, 2008.• Henrik Ohlsson, Jacob Roll, Anders Brun, Hans Knutsson, Mats Andersson and Lennart Ljung. Direct Weight Optimization Applied to Discontinuous Functions. In Proceedings of the 47th IEEE Conference
on Decision and Control, 2008.• Henrik Ohlsson. Regression on Manifolds with Implications for System Identification, Licentiate Thesis no. 1382, 2008.
5Nonlinear System Identification Using EM
A new algorithm for nonlinear system identification based on
•EM algorithm
• Particle smoother
Computed using a particle smoother
Composition parameters as a general representation of mixture components
A DAE Formulation For Multi-Zone Thermodynamic Models and It’s Application to CVCP Engines
• A DAE formulation for multi-zone in-cylinder models is extended with change of composition and the possibility to handle liquid components.
• The concept of composition parametersis introduced and it is shown that using these the same set of equations can be used to model the two fundamentally different cases of a Well Stirred Mixerand Well Stirred Reactor.
• It is shown that for all Well Behaved Gases the DAE has a unique solution for all pressure, temperatures, and compositions as long as all zones have non zero mass.
Addition of liquid zone
Extension with change of composition
Torque Modeling for Optimising Fuel Economy in Variable Compression Engines.Ylva Nilsson, Lars Eriksson, and Martin Gunnarsson (2008)International Journal of Modeling, Identification and Control (IJMIC), vol 3(no 3).
Fuel optimal control of a variable compression engine is studied and it is shown that a crucial component is the model for the engine torque.
The modeling and validation is performed on a VCR using two fuels with different octane ratings.
A model for the produced work that captures the important effects of ignition and compression ratio is proposed. The main task for the model is to be a mean for determining the fuel optimal control signals, for each requested engine torque and speed.
The contribution is a model suitable for finding this optimal combination. This model consists of well-known components, and the novelty lies in the compilation and validation of the control oriented efficiency model for a variable compression engine.
Despite the models simplicity, it describes the indicated work with good accuracy, and suits its purpose of finding optimal control signals.
In the evaluation it is shown that a fuel optimal controller based on the proposed model captures the optimal IMEP to within 1.2 % This corresponds to a loss in engine efficiency that is in the range of 0.5 percentage units or less.
Model Engine
Fault diagnosis
Structural analysis is a tool to identify submodelssuitable for design of diagnostic tests
The conventional structural analysis is extended to nonlinear models with inequality constraints
New tests are designed that can improve fault isolability properties of the diagnostic system
Åslund, J., and Frisk, E. Structural analysis for fault diagnosis of models with constraintsSafeProcess Symposium, Barcelona, June 2009
We can see where you are looking from measurements of your brain activity!
14 M
ay-0
9
Lots of signals Where you are lookingReal-time fMRI
• Henrik Ohlsson, Joakim Rydell, Anders Brun, Jacob Roll, Mats Andersson, Anders Ynnerman and Hans Knutsson. Enabling Bio-Feedback Using Real-Time fMRI. In
Proceedings of the 47th IEEE Conference on Decision and Control, 2008.
left
rightt t
Using Real-Time fMRI to Control a Dynamical System
Accepted at the conferenceInternational Society for Magnetic Resonance in Medicine
(ISMRM), Honolulu, April 2009Anders Eklund, Henrik Ohlsson, Mats Andersson, Joakim Rydell, Anders Ynnerman, Hans Knutsson
Brain computer interfaceA neural network classifies the brain activity every second
The classification is used as a force on the inverted pendulumThe subject balanced the inverted pendulum during 7 minutes
Using Real-Time fMRI to Control a Dynamical System
by Brain Activity ClassificationSubmitted to the conference
Medical Image Computing and Computer Assisted Intervention (MICCAI)
Anders Eklund, Henrik Ohlsson, Mats Andersson, Joakim Rydell, Anders Ynnerman, Hans Knutsson
Longer version of ISMRM abstractThe subject could balance the pendulum with imagined activity
Concurrent Volume Visualization of Real-Time fMRI
Nguyen Tan Khoa, Henrik Ohlsson, Anders Eklund, Willem J. Frishert, Patric Ljung, Mats Andersson,
Hans Knutsson, and Anders Ynnerman
Submitted to IEEE Visualization 2009
Clinical Applications for Real-Time fMRISubmitted to the conference
Medical Image Computing and Computer Assisted Intervention (MICCAI)
Anders Eklund, Nguyen Tan Khoa, Willem Frishert, Henrik Ohlsson, Mats Andersson, Anders Ynnerman, Hans Knutsson
Interactive functional brain mapping as surgical aidBio-feedback, e.g. learn to suppress your pain
Stroke and Parkinson rehabilitation
Free Form Incident Light FieldsJonas Unger, Stefan Gustavson, Per Larsson and Anders Ynnerman
VITA, Linköping University
Computer Graphics Forum Vol. 27 Nr. 4, 2008The Eurographics Association and Blackwell Publishing Ltd.
Special issue EGSR '08,
Capture Processing Rendering
HDR image stream
Tracking information
ILF scene description
Extracted illuminantsBackground ILF
Free Form Incident Light Fields
• A technique for photo-realistic rendering using captured real world illumination
• Illumination capture using High Dynamic Range video imaging
• Projection onto a set of 2D surfaces in the scene
• A novel data structure for efficient random access of the 4D radiance data suitable for image synthesis
• Algorithm for photo-realistic rendering of synthetic objects placed in the captured real world scene
z
x
y
HDR Light Probe Sequence Resampling for Realtime Incident
Light Field RenderingJoakim Löw, Anders Ynnerman, Per Larsson, Jonas Unger
VITALinköping University
● Image Based Lighting– HDR measurements of
real world illumination
● Resampled to grid of spherical harmonics– Realtime lighting
– Memory efficient
HDR Light Probe Sequence Resampling for Realtime Incident Light Field Rendering
The Impact of Feedback Designon Haptic Volume Visualization
Karljohan Lundin Palmerius Camilla Forsell
Norrköping Visualization and Interaction StudioLinköping University, Sweden
Dynamic Deformation Using Adaptable, Dynamic Deformation Using Adaptable, Linked Asynchronous FEM RegionsLinked Asynchronous FEM Regions
UmutUmut KocakKocak, , KarljohanKarljohan LundinLundin PalmeriusPalmerius, Matthew Cooper, Matthew Cooper
MOVIII workshop2009-05-14
Prof Fredrik GustafssonDr Thomas SchönDr Umut OrgunerDr Gustaf HendebyDr David TörnqvistLic Jeroen HolKarl GranströmJonas CallmerFredrik Lindsten
MOVIII: Information Integration Publications 2008
MOVIII workshop2009-05-14
Plenary presentation
Fusion of data from different sources
Author: T. Schön
Published: Proceeding of the 10th International Symposium on 3D Analysis of Human Movement (3DMA), 2008.
The H.J. Woltring lecture at 3DMA
MOVIII workshop2009-05-14
Two new courses
Two courses using MOVIII data/core problems
Graduate course: Dynamic vision, Poster
Undergraduate course: Sensor fusion, Poster
SLAM LAB
Rotation invariant visual odometry project
MOVIII workshop2009-05-14
Theses
Jeroen Hol, licentiate thesis: Pose estimation and calibration algorithms for vision and inertial sensors
David Törnqvist, PhD thesis: Estimation and detection with applications to navigation
MOVIII workshop2009-05-14
Particle filter theory
A basic convergence result for particle filtering
Authors: X. L. Hu, T. Schön, L. Ljung Published: IEEE Transactions on Signal Processing, Vol 56, pp
1337-1348, 2008.
Risk-Sensitive Particle Filters for Mitigating Sample Impoverishment Authors: U. Orguner, F. Gustafsson
Published: IEEE Transactions on Signal Processing, Vol 56, 2008.
Journal papers:• A new convergence proof of the PF with stronger results (the state estimate converges, not only bounded functions of the state).• An new alternative design and interpretation of jitter noise in the PF in terms of risk minimization.
MOVIII workshop2009-05-14
Particle smoothing in system ident.
Particle Filter Approach to Nonlinear System Identi¯ cation under Missing Observations with a Real Application
Authors: R. B. Gopaluni, T. Schön, A. G. Wills
Published: Proceedings of the 15th IFAC Symposium on System Identification (SYSID), 2009.
Parameter Estimation for Discrete-Time Nonlinear Systems Using EM
Authors: A. Wills, T. Schön, B. Ninness
Published: Proceedings of the 17th IFAC World Congress, 2008.
MOVIII workshop2009-05-14
Target tracking
Improved Target Tracking with Road Network Information Authors: U. Orguner, T. Schön, F. Gustafsson
Published: Proceedings of IEEE Aerospace, 2009. POSTERTarget tracking using delayed measurements with implicit constraints Authors: U. Orguner, F. Gustafsson
Published: Proceedings of 11th International Conference on Information Fusion, 2008.One of five papers selected for best paper awardStorage efficient particle filters for the out of sequence measurement problem
Authors: U. Orguner, F. Gustafsson Published: Proceedings of 11th International Conference on Information Fusion, 2008.
Illustration: Propagation delayed measurements give state dependent time delay in measurement equation
MOVIII workshop2009-05-14
Collision avoidance
IFAC symposium Safeprocess, Barcelona, 2009.
Authors: Fredrik Lindsten, Per-Johan Nordlund, Fredrik Gustafsson
Poster
MOVIII workshop2009-05-14
Calibration of camera and IMU units
A New Algorithm for Calibrating a Combined Camera and IMU Sensor Unit
Authors: J. Hol, T. Schön, F. Gustafsson
Published: Proceedings of the 10th International Conference on Control, Automation, Robotics and Vision (ICARCV), 2008.
Relative Pose Calibration of a Spherical Camera and an IMU
Authors: J. Hol, T. Schön, F. Gustafsson
Published: Proceedings of the 7th IEEE and ACM International Symposium on Mixed and Augmented Reality (ISMAR), 2008.
MOVIII workshop2009-05-14
Other camera related issues
Detecting Spurious Features using Parity Space
Authors: D. Törnqvist, T. Schön, F. Gustafsson Published: International Conference on Control,
Automation, Robotics and Vision, 2008.
Separate presentation
Indoor photorealistic 3D mapping using stereo images from SLR cameras
Authors: V. Kolbe, F. Isaksson, T. Beckman, T. Schön
Published: Swedish Symposium on Image Analysis, 2009.
Awarded second prize for best industrial contribution
Ego-Motion and Indirect Road Geometry Estimation Using Night Vision
Authors: T. Schön, J. Roll Published: Proceedings of the IEEE Intelligent
Vehicle Symposium (IV), 2009.
MOVIII workshop2009-05-14
SLAM
Utilizing Model Structure for Efficient Simultaneous Localization and Mapping for a UAV Application
Authors: R. Karlsson, T. Schön, D. Törnqvist, G. Conte, F. Gustafsson Published: Proceedings 2008 IEEE Aerospace Conference, 2008.
Tree of Words for Visual Loop Closure Detection in Urban SLAM Authors: J. Callmer, K. Granström, J. Nieto, F. Ramos
Published: Proceedings of the 2008 Australasian Conference on Robotics & Automation, pp 8, 2008.
Learning to detect loop closure from range data Authors: Callmer, J.; Granstrom, K.; Nieto, J.I.; Ramos, F.T
Published: Proceedings of the 2009 IEEE International Conference on Robotics and Automation (ICRA)
MOVIII workshop2009-05-14
UT/UKF
On Nonlinear Transformations of Stochastic Variables and its Application to Nonlinear Filtering
Authors: G. Hendeby, F. Gustafsson
Published: International Conference on Acoustics, Speech, and Signal Processing (ICASSP), 2008.
Knowledge Processing Middleware
■ Principled and systematic software framework for bridging the sense-reasoning gap in a physical agent
UAV State Estimation
Image Processing
Qualitative Spatial Reasoning
Chronicle Recognition
GPS
Pan-tilt unit
Color camera
Thermal camera
IMU
Anchoring
Heintz, Kvarnström and Doherty. Knowledge Processing Middleware. 1st International Conference on Simulation, Modeling and Programming of Autonomous Robots (SIMPAR), 2008.
Heintz, Kvarnström and Doherty. Bridging the Sense-Reasoning Gap: DyKnow – A Middleware Framework for Knowledge Processing. IROS Workshop on Current Software Frameworks in Cognitive Robotics Integrating Different Computational Paradigms, 2008.
Qualitative models
Knowledgefusion
Mixed models
Informationfusion
Quantitative models
Sensor data
Sensorfusion
DyKnow Federations
■ DyKnow supports information processing within a single UAV■ A DyKnow federation allows UAVs to share and merge
streams of information on many levels of abstraction■ Example: Detecting traffic situations using multiple UAVs
Platform 1
Gatewayagent
expo
rtpr
oxy
impo
rtpr
oxy
FIPA ACL
Directcommunication
Interfaceagent
DyKnowfederation
agent
Efficient streaming,low latency
Set up connectionusing speech acts
Platform 2
Gatewayagent
expo
rtpr
oxy
impo
rtpr
oxy
Interfaceagent
DyKnowfederation
agent
Heintz and Doherty. DyKnow Federations: Distributing and Merging Information Among UAVs.11th International Conference on Information Fusion, 2008.
Planning and Execution Monitoring
■ We use automated planning to achieve complex goals■ Plans may fail■ Monitor plan execution!
■ DyKnow generates state sequences■ Constrained by temporal logic formulas■ operator attach(uav, crate)
:monitor executing-attach(uav, crate) until [0,5000](carrying(uav, crate) until executing-detach(uav, crate))
Doherty, Kvarnström and Heintz. A Temporal Logic-based Planning and Execution Monitoring Framework for Unmanned Aircraft Systems.Journal of Autonomous Agents and Multi-Agent Systems (JAAMAS), 2009.
Kvarnström, Heintz and Doherty. A Temporal Logic-Based Planning and Execution Monitoring System. 18th International Conference on Automated Planning and Scheduling (ICAPS), 2008.
Fly Attach Detachcarrying(uav,crate)
…
at most 5000 ms
Stream Reasoning in DyKnow
■ DyKnow represents information about the world as streams.■ By incrementally reasoning about the world as new information becomes
available it is possible to draw conclusions with minimal delay.■ Two examples of stream reasoning in DyKnow:
■ Evaluating temporal logical formulas over streams of states using progression. executing-attach(uav, crate) until [0,5000] (carrying(uav, crate) until executing-detach(uav, crate))
■ Detecting complex situations described by simple temporal networks of events over streams of event occurrences using chronicle recognition.
Doherty, Kvarnström and Heintz. Stream Reasoning in DyKnow: A Knowledge Processing Middleware System. ECSW Workshop on Stream Reasoning (SR), 2009.
A B C
D E
[0, 2000] [500, 1000][0, 1000]
[1000, 1000][1000, 2000]
A Reconfigurable Diagnosis System
■ Problem formulation:■ multiple fault diagnosis (MFD)■ real-time requirements■ dynamic systems
■ Solution:■ tests with thresholded residuals■ accurate MFD requires many tests,
however all tests are not needed at all times
■ we propose FlexDx, a reconfigurable diagnosis system only running testsneeded for the current diagnoses which starts tests in the past if necessary
Heintz, Krysander, Roll and Frisk. FlexDx: A reconfigurable diagnosis framework. 19th International Workshop on Principles of Diagnosis (DX), 2008.
Krysander, Heintz, Roll and Frisk. Dynamic test selection for reconfigurable diagnosis. 47th IEEE Conference on Decision and Control (CDC), 2008.
■ Key contributions:■ How to select and initiate tests
given the current diagnoses, and
■ how to implement FlexDx using the stream-based DyKnow middleware.
TestTestResidualResidual
DiagnosesTestSet
System
system observations residuals
last fault free time
test set
diagnoses initial diagnosis
conflict set
ResidualGenerator
ConflictSetMediator
CreateTests
ResidualMonitor
Positioning Communication Relays
■ UAVs used for monitoring of important targets■ High volume transmission requires line-of-sight■ Problematic in urban or mountainous terrain
■ Use chain of intermediate relay UAVs – must find suitable placement!■ Two new algorithms developed, based on discretization
■ Experiments: Faster by a factor of 10–50
Burdakov, Doherty, Holmberg, Kvarnström and Olsson. Positioning Unmanned Aerial Vehicles as Communication Relays for Surveillance Tasks. Robotics: Science and Systems (RSS), 2009.
Burdakov, Doherty, Holmberg and Olsson. Optimal placement of communications relay nodes.23rd European Conference on Operational Research (EURO), 2009.
Logical Reasoning
■ Logical reasoning is powerful, in theory■ We want to make it practical
■ FCTP algorithm for logical reasoning with assumptions■ Sound and complete for a well-defined class of theories■ Fully interruptible
Magnusson, Kvarnström and Doherty. Abductive Reasoning with Filtered Circumscription. 8th Workshop on Nonmonotonic Reasoning, Action and Change (NRAC), 2009.
FCTP
Communication
■ Complex real-world scenarios involve cooperation■ Cooperation requires communication
■ Logical reasoning about belief and commitment■ Planning speech acts and cooperation
Magnusson and Doherty. Planning Speech Acts in a Logic of Action and Change.25th Workshop of the Swedish Artificial Intelligence Society (SAIS), 2009.
Execution
■ Plans break down when executed
■ Fully automatic execution monitoring■ Fully automatic plan revision
Magnusson, Landén and Doherty. Planning, Executing, and Monitoring Communication in a Logic-based Multi-agent System. 18th European Conference on Artificial Intelligence (ECAI), 2008.
Magnusson, Landén and Doherty. Logical Agents that Plan, Execute, and Monitor Communication. 2nd Workshop on Logic and the Simulation of Interaction and Reasoning (LSIR-2), 2009.
Interactivity
■ We have autonomous cooperating robot agents■ Now if only we could communicate with them!
■ Natural language question answering■ Interactive computer game application
Magnusson and Doherty. Logical Agents for Language and Action.4th Conference on Artificial Intelligence and Interactive Digital Entertainment (AIIDE), 2008.
Who are you?
I am Smith.
Loops
■ Logical reasoning is powerful, in theory■ We’ve made it (more) practical■ But is it still powerful?
■ Automated induction proofs■ Infinite UAV surveillance plans■ Compact representation using fixpoint logic
Magnusson and Doherty. Deductive Planning with Inductive Loops.11th International Conference on Principles of Knowledge Representation and Reasoning (KR), 2008.
Non-GPS Navigation
■ The outcome of this research is a navigation system which combines complementary vision-based techniques together with inertial sensors to navigate a UAV in case of GPS outages
■ The approach proposed relies on geo-referenced imagery information stored in the platform and it is based on probabilistic techniques
■ The navigation system developed has been implemented and tested on-board an experimental unmanned helicopter platform
Conte and Doherty. Vision-based Unmanned Aerial Vehicle Navigation Using Geo-referenced Information. Accepted for publication: EURASIP Journal of Advances in Signal Processing, 2009.
Vision-based Indoor Navigation
A method for accurate vision-based state estimation which allows for autonomous indoor navigation of micro-scale UAS’s has been implemented and tested in real-flight closed-loop experiments.
The system consists of vision-based pose estimation, sensor fusion by means of a 9 state Kalman filter, and control executed onboard the UAS. Autonomous flight capability is achieved without communication with external entities.