SlideShare uma empresa Scribd logo
1 de 17
Baixar para ler offline
Optimizing the configuration
 of X-by-Wire networks using
     word combinatorics
         Nicolas NAVET - Bruno GAUJAL
           LORIA Lab. (Nancy, France)
                    TRIO Team
            http://www.loria.fr/~nnavet

                 EPFL Seminar
             Network Calculus Group




TTP/C – Time Triggered Protocol

 Designed at T.U. Vienna + TTTech
 TTP/C main technical characteristics:
 •   Determinism
 •   Fault-Tolerance
 •   Composability
 •   Support of mode changes

     A good candidate for X-By-Wire ..


                  N. NAVET - EPFL - 02/03/2004   2
X-by-Wire
  Hydraulic and mechanical connection are
  replaced by networks and actuators

  Why ?
    Decrease of weight and cost
    Safety : intrusion of the steering column in the
    cockpit
    New functions : variable demultiplication - crash
    avoidance
    Less pollution (brake / transmission liquid)
    …
                 N. NAVET - EPFL - 02/03/2004           3




X-by-wire : an example




                                « Steer-by-wire »
                 N. NAVET - EPFL - 02/03/2004           4
A TTP/C Cluster

Small Replaceable
   Unit (SRU)




 Medium Access Control : TDMA
 Redundant transmission support
 Data rate: 500kbit/s, 1Mbit/s, 2Mbit/s, 5Mbit/s, 25Mbit/s
 Topology: bus or star
                    N. NAVET - EPFL - 02/03/2004          5




TDMA – Time division Multiplexed Access




Slot: time window given to a station for a transmission
TDMA Round: sequence of slots s.t. each station
transmits exactly once
Cluster Cycle: sequence of the              TDMA rounds

                    N. NAVET - EPFL - 02/03/2004          6
TTP/C: Implications of the MAC protocol
Bounded response times and « heartbeats » but:
  loss of bandwidth
  need of powerful CPU’s
  maximum timing contraint:
      If a station sends a single information, the refresh cannot
      be more frequent than the length of a round
      If a station sends several informations, the refresh cannot
      be more frequent than 2x the length of a round
 Ex: 5ms time constraint - 500kbit/s network with 200 bits per
 frames - at most 12 frames (6 FTUs of two nodes) or 6 frames
 if the station sends 2 distinct informations


                     N. NAVET - EPFL - 02/03/2004                     7




FTU: Fault Tolerant Unit

  FTU = set of stations that act identically




    FTU of cardinality 2                 FTU of cardinality 3
                                        (majority vote is possible)
  Replica = a frame sent by a node of the FTU
                     N. NAVET - EPFL - 02/03/2004                     8
FTU: which protection ?
  Protection against:
     disappearance of a station (crash, disconnection..)
     corrupted frames (EMI)
     sensors or computation errors
     …
  Under the assumption of a single failure (TTP/C
  fault-hypothesis) :
     A dual redundancy ensures a protection in « the temporal
     domain »
     A triple redundancy ensures in addition a protection in « the
     value domain »
  Problem: history-state
                        N. NAVET - EPFL - 02/03/2004                  9




Goal of the study: maximize the
robustness against transmission errors
   Transmission errors are usually highly correlated

    Fault         Fault
   Tolerant
    Unit A
                 Tolerant
                  Unit B
                                              Where to
                                             place the
                ECU2
   ECUA
   ECUA12
                ECU2
               ECUB1                         replicas ?
                   b
   a1 a2      b1 b2 3
                                      a1 b1 a2 b2 b3 a1 b1 a2 b2 b3   …
TTP/C Bus
                                                                           t
                                      TTP Round



                        N. NAVET - EPFL - 02/03/2004                  10
Application model
  Ta : production cycle of the data sent by the
  stations of the FTU a
                                             Ta
          a1   a2        a3      a1     a2          a3   a1   a2   a3

         TTP Round

  Assumptions:
   • no synchronization between production and
     transmission (round)
   • production cycle is a multiple of the length of a
     round
                     N. NAVET - EPFL - 02/03/2004                  11




Objective w.r.t. fail-silence
  A node is « fail-silent » if one can safely
  consume its data when the frame carrying the
  data is syntaxically correct

  Stations are fail-silent: « minimize Pall :
  the probability that all frames of the FTU sent
  during a production cycle will be corrupted »

  Stations are not fail-silent: « minimize
  Pone : the probability that at least one frame of
  the FTU will be corrupted»

                     N. NAVET - EPFL - 02/03/2004                  12
Assumptions on the error model
  Each bit transmitted during an EMI will be corrupted
  with probability
  If a perturbation overlaps a whole slot, the
  corresponding frame is corrupted with probability 1
  Starting times of the EMI bursts are independent and
  uniformly distributed over time
  The distribution of the size of the bursts is arbitrary

                A C                B
           a1   a2        a3      a1    a2          a3
                                                         t
          TTP Round

                     N. NAVET - EPFL - 02/03/2004            13




  Objective 1 : Minimize Pone
Majorization - Schur-Convexity
  vector u          (u1 ,..., un ) majorizes v                        ( v1 ,..., vn ) if:
                n           n                   k               k
                      ui         vi and               u[i ]           v[i ] k       n
                i 1        i 1                  i 1             i 1

     with (u[ i ] ,..., u[ n ] ) permutation of u s.t. u[i ]                             ... u[ n ]

  Example: (1,3,5,10)                     (2, 4, 4,9)
                           n
  A fonction          f:                  is
    Schur-convex if                   u     v          f (u )          f (v )
    Schur-concave if u                      v          f (u )          f (v )
                           N. NAVET - EPFL - 02/03/2004                                         15




Minimizing Pone
  I( x ) is the vector of the distance between replicas
  during the length of a round (sorted in ascending
  order)

  Example: I( x )               (1,1, 2)
           a1         a2         a3        a1       a2              a3
                                                                                    t
          TTP round
   Example: I( x )              (0,0, 4)
        a1 a2 a3                      a1 a2 a3
                                                                                     t
           TTP round


                           N. NAVET - EPFL - 02/03/2004                                         16
Minimizing Pone

Theorem: the best allocation for Pone is to group
  together all replicas (denoted allocation g)

Arguments:
• Pone is shur-concave: I( x ') I( x )       Pone ( x ') Pone ( x )
• I( g ) is maximum for the majorization (equal to
  (0,0,..., S k ) with k the number of replicas of the FTU
   and S the number of slots per round)




                          N. NAVET - EPFL - 02/03/2004                  17




Minimize Pone
  Idea of the proof (step 1): the farther the beginning of
  an error burst from a replica, the less likely the replica
  becomes corrupted. « Non-grouped » allocations
  have more areas close to replicas
       a1   a2       a3       a1    a2          a3       allocation x
                                                              t



    a1 a2 a3              a1 a2 a3                       allocation g
                                                               t

                 P( x) P( g)                P( x) P( g)

                          N. NAVET - EPFL - 02/03/2004                  18
Minimize Pone

 Validity of the result :
  • Arbitrary value and burst size distribution
  • Production period multiple of the round length
  • for all TDMA networks

 Combined minimization of Pone for all FTU’s
 is possible
 Robustness improvement: against a random
 allocation, the number of lost data is reduced from 15 to
 20% on average

                    N. NAVET - EPFL - 02/03/2004        19




   Objective 2 : Minimize Pall
          TTP/C case
TTP/C : the majority rule
    Cliques: sets of stations that disagree on the state of the
    network
    Principle: to avoid cliques, stations in the minority
    disconnect (« freeze »)
    Mechanism: before sending, a station checks that in the
    last round (S slots), the number of correct messages is
    greater than the number of incorrect messages,
    otherwise it disconnects
                 ai-1                     ai
                                                        t
               S slots in a round
     If a station « freezes » due to transmission errors, the
     others follow one by one...
                        N. NAVET - EPFL - 02/03/2004        21




 TTP/C : minimize Pall
Algorithm:    1) for each FTU i with Ci slots, push Ci / 2
                 slots in the smallest stack and Ci / 2 in the
                 largest stack
               2) concatenate the two stacks
Ex: FTU A: 3 replicas – FTU B: 2 replicas – FTU C: 4 replicas




                        stack 1            stack 2
    TTP/C round :
                        N. NAVET - EPFL - 02/03/2004        22
TTP/C : minimize Pall
Theorem: the « 2-stacks » algorithm is optimal
  under TTP/C

Arguments:
  Case 1) a perturbation for each replica : identical  allocation
  Case 2) a perturbation can corrupt several replicas with a
  probability decreasing in the distance between the replicas. A
  burst of more than S / 2 slots freezes the system, now the
  algorithm ensures a distance of S / 2 slots

Corollary: it is useless to have more than 2 replicas
  per FTU if the probability to have more than one
  perturbation in the same round is sufficiently low
                      N. NAVET - EPFL - 02/03/2004          23




    Objective 2 : Minimize Pall
           TDMA case
Balanced words
A « balanced » word (or Sturmian word) is a binary
sequence un n s.t. :
                                     n k         m k
              k , n, m                     ui          uj       1
                                     i n         j m

Balanced words are computed using bracket sequences :
                                     a                      a
                     un          n              n 1
                                     b                      b
 where a / b is the rate of the word (nb of 1 / nb of 0)
Example: balanced word of rate 3/8
                           (0,0,1,0,0,1,0,1)
                          N. NAVET - EPFL - 02/03/2004                   25




Multimodular functions
Multimodularity [Hajek] : counterpart of convexity for
discrete functions f : m

Definition : Let F be a set of m+1-vectors that sum to
                             m
   0, a function f :                       is F-multimodular if
          f ( x v)        f ( x w)              f ( x)      f ( x v w)
               m
         x         et v, w F , v                w

 Example: x (0,1,0,...,1,1,0) is a control sequence,
  f a cost function and v an elementary operation
 moving a client to the left
                      v       (0,...,0,1, 1,0,...,0)
                          N. NAVET - EPFL - 02/03/2004                   26
Optimisation and multimodular function
  Global left shift operator : si ( x )
                ex: s2 (0,1,0,1,1,0)            (0,1,1,0,0,1)

  Theorem [Altman,Gaujal,Hordijk 97]: If f is multimodular
                    m
       then G ( x ) 1/ m i 1 f ( si ( x )) (shift-invariant version
       of f ) is minimum (among all admissible sequences) if x
       is a balanced sequence.

  Theorem : If the size of the bursts is exponentially
    distributed then Pall is multimodular.
  Moreover, Pall is equal to its shift invariant version thus
    Pall is minimum for a balanced sequence.

                          N. NAVET - EPFL - 02/03/2004          27




 Optimal algorithm : a single FTU
  FTU with C replicas of size h bits in a round of total
  size R bits
       compute vi balanced word of rate C /( R C ( h 1))
        x is the round initially empty
       If vi 1 then x : x 1...1 (h '1' concatenated)
       If vi 0 then x : x 0
Ex: FTU: 3 replicas of cardinality 3 in a round of size 14
  vi     (0,0,1,0,0,1,0,1) with rate 3/ 8
  x     (0,0,1,1,1,0,0,1,1,1,0,1,1,1) ( balanced word with rate 9 /14


                          N. NAVET - EPFL - 02/03/2004          28
Optimal algorithm : several FTUs
Problem : allocation conflicts
Ex: FTU A: 3 replicas – FTU B: 2 replicas – FTU C: 1 replica
              x A (0,1,0,1,0,1) rate 3/ 6
              xB (0,0,1,0,0,1) rate 2 / 6
              xC (0,0,0,0,0,1) rate 1/ 6
An optimal allocation is still possible if
 • the number of cardinalities is a power of 2
 • all replicas have the same size

Remark: a balanced sequence is minimum for the
majorization order, it is thus the worst solution for
Pone! Both objectives are contradictory
                     N. NAVET - EPFL - 02/03/2004       29




Heuristic : several FTUs

The rate of emission: number of bits that FTU A
must emit on average during each bit of a round :
                         dA      C Ah A / R


At step i, one schedules the transmission of a
frame of the FTU for which the number of due bits
– number of already allocated bits is maximum




                     N. NAVET - EPFL - 02/03/2004       30
Pall : Heuristic vs random allocation
 Reduction of the number of lost messages
 w.r.t. a random allocation :




                 N. NAVET - EPFL - 02/03/2004        31




Pall : Heuristic vs optimal
 Increase of the number of lost messages w.r.t. to
 the optimal :




                 N. NAVET - EPFL - 02/03/2004        32
Conclusion
Optimal and near optimal allocation policies for TDMA
 and TTP/C networks

  Choice of the locations of the slots have a strong
  influence on the robustness of the network
  The cost function plays a major role on the shape
  of the solution
  Hypothesis on the error model are crucial

Future work :
  Configurations made of fail-silent and non fail-silent
  nodes (minimizing Pone and Pall for different FTU’s)
  FlexRay protocol
                    N. NAVET - EPFL - 02/03/2004       33

Mais conteúdo relacionado

Mais procurados

TVWS_NGMAST2015
TVWS_NGMAST2015TVWS_NGMAST2015
TVWS_NGMAST2015
mmlodro
 
Time reversed acoustics - Mathias Fink
Time reversed acoustics - Mathias FinkTime reversed acoustics - Mathias Fink
Time reversed acoustics - Mathias Fink
Sébastien Popoff
 

Mais procurados (20)

DSP_2018_FOEHU - Lec 06 - FIR Filter Design
DSP_2018_FOEHU - Lec 06 - FIR Filter DesignDSP_2018_FOEHU - Lec 06 - FIR Filter Design
DSP_2018_FOEHU - Lec 06 - FIR Filter Design
 
DSP_2018_FOEHU - Lec 02 - Sampling of Continuous Time Signals
DSP_2018_FOEHU - Lec 02 - Sampling of Continuous Time SignalsDSP_2018_FOEHU - Lec 02 - Sampling of Continuous Time Signals
DSP_2018_FOEHU - Lec 02 - Sampling of Continuous Time Signals
 
A Simple Communication System Design Lab #1 with MATLAB Simulink
A Simple Communication System Design Lab #1 with MATLAB Simulink A Simple Communication System Design Lab #1 with MATLAB Simulink
A Simple Communication System Design Lab #1 with MATLAB Simulink
 
TVWS_NGMAST2015
TVWS_NGMAST2015TVWS_NGMAST2015
TVWS_NGMAST2015
 
Dcp project
Dcp projectDcp project
Dcp project
 
FFT
FFTFFT
FFT
 
Sampling and Reconstruction of Signal using Aliasing
Sampling and Reconstruction of Signal using AliasingSampling and Reconstruction of Signal using Aliasing
Sampling and Reconstruction of Signal using Aliasing
 
Software-defined white-space cognitive systems: implementation of the spectru...
Software-defined white-space cognitive systems: implementation of the spectru...Software-defined white-space cognitive systems: implementation of the spectru...
Software-defined white-space cognitive systems: implementation of the spectru...
 
10 range and doppler measurements in radar systems
10 range and doppler measurements in radar systems10 range and doppler measurements in radar systems
10 range and doppler measurements in radar systems
 
Chapter 2
Chapter 2Chapter 2
Chapter 2
 
Matlab code for comparing two microphone files
Matlab code for comparing two microphone filesMatlab code for comparing two microphone files
Matlab code for comparing two microphone files
 
Fast Fourier Transform
Fast Fourier TransformFast Fourier Transform
Fast Fourier Transform
 
Fft analysis
Fft analysisFft analysis
Fft analysis
 
Lecture13
Lecture13Lecture13
Lecture13
 
Matlab implementation of fast fourier transform
Matlab implementation of  fast fourier transformMatlab implementation of  fast fourier transform
Matlab implementation of fast fourier transform
 
Dereverberation in the stft and log mel frequency feature domains
Dereverberation in the stft and log mel frequency feature domainsDereverberation in the stft and log mel frequency feature domains
Dereverberation in the stft and log mel frequency feature domains
 
Fft analysis
Fft analysisFft analysis
Fft analysis
 
Design and Implementation of Low Ripple Low Power Digital Phase-Locked Loop
Design and Implementation of Low Ripple Low Power Digital Phase-Locked LoopDesign and Implementation of Low Ripple Low Power Digital Phase-Locked Loop
Design and Implementation of Low Ripple Low Power Digital Phase-Locked Loop
 
Time reversed acoustics - Mathias Fink
Time reversed acoustics - Mathias FinkTime reversed acoustics - Mathias Fink
Time reversed acoustics - Mathias Fink
 
Modulation techniques matlab_code
Modulation techniques matlab_codeModulation techniques matlab_code
Modulation techniques matlab_code
 

Destaque

Ertss2010 multicore scheduling
Ertss2010 multicore schedulingErtss2010 multicore scheduling
Ertss2010 multicore scheduling
Nicolas Navet
 
Join The Festivities By Clicking LIKE..
Join The Festivities By Clicking LIKE..Join The Festivities By Clicking LIKE..
Join The Festivities By Clicking LIKE..
chalicesnowflower
 
Cief2007 nn shc_slides
Cief2007 nn shc_slidesCief2007 nn shc_slides
Cief2007 nn shc_slides
Nicolas Navet
 

Destaque (15)

Automotive communication systems: from dependability to security
Automotive communication systems: from dependability to securityAutomotive communication systems: from dependability to security
Automotive communication systems: from dependability to security
 
Isi2007 nn shc_2007
Isi2007 nn shc_2007Isi2007 nn shc_2007
Isi2007 nn shc_2007
 
Ertss2010 multicore scheduling
Ertss2010 multicore schedulingErtss2010 multicore scheduling
Ertss2010 multicore scheduling
 
New policies
New policiesNew policies
New policies
 
Pushing the limits of CAN - Scheduling frames with offsets provides a major p...
Pushing the limits of CAN - Scheduling frames with offsets provides a major p...Pushing the limits of CAN - Scheduling frames with offsets provides a major p...
Pushing the limits of CAN - Scheduling frames with offsets provides a major p...
 
Excel Basics
Excel BasicsExcel Basics
Excel Basics
 
Virtualization in Automotive Embedded Systems : an Outlook
Virtualization in Automotive Embedded Systems : an OutlookVirtualization in Automotive Embedded Systems : an Outlook
Virtualization in Automotive Embedded Systems : an Outlook
 
Join The Festivities By Clicking LIKE..
Join The Festivities By Clicking LIKE..Join The Festivities By Clicking LIKE..
Join The Festivities By Clicking LIKE..
 
Aperiodic Traffic in Response Time Analyses with Adjustable Safety Level
Aperiodic Traffic in Response Time Analyses with Adjustable Safety LevelAperiodic Traffic in Response Time Analyses with Adjustable Safety Level
Aperiodic Traffic in Response Time Analyses with Adjustable Safety Level
 
In-Vehicle Networking : a Survey and Look Forward
In-Vehicle Networking : a Survey and Look ForwardIn-Vehicle Networking : a Survey and Look Forward
In-Vehicle Networking : a Survey and Look Forward
 
Cief2007 nn shc_slides
Cief2007 nn shc_slidesCief2007 nn shc_slides
Cief2007 nn shc_slides
 
Battery Aware Dynamic Scheduling for Periodic Task Graphs
Battery Aware Dynamic Scheduling for Periodic Task GraphsBattery Aware Dynamic Scheduling for Periodic Task Graphs
Battery Aware Dynamic Scheduling for Periodic Task Graphs
 
Automating the Configuration of the FlexRay Communication Cycle
Automating the Configuration of the FlexRay Communication CycleAutomating the Configuration of the FlexRay Communication Cycle
Automating the Configuration of the FlexRay Communication Cycle
 
Configuring the communication on FlexRay: the case of the static segment
Configuring the communication on FlexRay: the case of the static segmentConfiguring the communication on FlexRay: the case of the static segment
Configuring the communication on FlexRay: the case of the static segment
 
Programació RAC1 2012-13
Programació RAC1 2012-13Programació RAC1 2012-13
Programació RAC1 2012-13
 

Semelhante a Optimizing the Robustness of X-by-Wire using Word Combinatorics

Design and Fabrication of a Two Axis Parabolic Solar Dish Collector
Design and Fabrication of a Two Axis Parabolic Solar Dish CollectorDesign and Fabrication of a Two Axis Parabolic Solar Dish Collector
Design and Fabrication of a Two Axis Parabolic Solar Dish Collector
IJERA Editor
 
CS Based Channel Estimation for OFDM Systems under Long Delay Channels Using ...
CS Based Channel Estimation for OFDM Systems under Long Delay Channels Using ...CS Based Channel Estimation for OFDM Systems under Long Delay Channels Using ...
CS Based Channel Estimation for OFDM Systems under Long Delay Channels Using ...
IJERA Editor
 
Iaetsd a novel scheduling algorithms for mimo based wireless networks
Iaetsd a novel scheduling algorithms for mimo based wireless networksIaetsd a novel scheduling algorithms for mimo based wireless networks
Iaetsd a novel scheduling algorithms for mimo based wireless networks
Iaetsd Iaetsd
 
03 time division-multiplexing
03 time division-multiplexing03 time division-multiplexing
03 time division-multiplexing
Rafeek Mohamed
 

Semelhante a Optimizing the Robustness of X-by-Wire using Word Combinatorics (20)

2015 08-31 kofidis
2015 08-31 kofidis2015 08-31 kofidis
2015 08-31 kofidis
 
Design and Fabrication of a Two Axis Parabolic Solar Dish Collector
Design and Fabrication of a Two Axis Parabolic Solar Dish CollectorDesign and Fabrication of a Two Axis Parabolic Solar Dish Collector
Design and Fabrication of a Two Axis Parabolic Solar Dish Collector
 
upload2
upload2upload2
upload2
 
A JOINT TIMING OFFSET AND CHANNEL ESTIMATION USING FRACTIONAL FOURIER TRANSFO...
A JOINT TIMING OFFSET AND CHANNEL ESTIMATION USING FRACTIONAL FOURIER TRANSFO...A JOINT TIMING OFFSET AND CHANNEL ESTIMATION USING FRACTIONAL FOURIER TRANSFO...
A JOINT TIMING OFFSET AND CHANNEL ESTIMATION USING FRACTIONAL FOURIER TRANSFO...
 
Mac layer
Mac  layerMac  layer
Mac layer
 
Multi carrier equalization by restoration of redundanc y (merry) for adaptive...
Multi carrier equalization by restoration of redundanc y (merry) for adaptive...Multi carrier equalization by restoration of redundanc y (merry) for adaptive...
Multi carrier equalization by restoration of redundanc y (merry) for adaptive...
 
ofdm
ofdmofdm
ofdm
 
Gprs
GprsGprs
Gprs
 
6 Switch Fabric
6 Switch Fabric6 Switch Fabric
6 Switch Fabric
 
IMT Advanced
IMT AdvancedIMT Advanced
IMT Advanced
 
Analysis of multipath channel delay estimation using subspace fitting
Analysis of multipath channel delay estimation using subspace fittingAnalysis of multipath channel delay estimation using subspace fitting
Analysis of multipath channel delay estimation using subspace fitting
 
Phd defense of xin you
Phd defense of xin youPhd defense of xin you
Phd defense of xin you
 
CS Based Channel Estimation for OFDM Systems under Long Delay Channels Using ...
CS Based Channel Estimation for OFDM Systems under Long Delay Channels Using ...CS Based Channel Estimation for OFDM Systems under Long Delay Channels Using ...
CS Based Channel Estimation for OFDM Systems under Long Delay Channels Using ...
 
13486500-FFT.ppt
13486500-FFT.ppt13486500-FFT.ppt
13486500-FFT.ppt
 
Frequency Modulation
Frequency ModulationFrequency Modulation
Frequency Modulation
 
Printed Omnidirectional Access-Point Antenna for 2.4/5-GHz WLAN Operation
Printed Omnidirectional Access-Point Antenna for 2.4/5-GHz WLAN OperationPrinted Omnidirectional Access-Point Antenna for 2.4/5-GHz WLAN Operation
Printed Omnidirectional Access-Point Antenna for 2.4/5-GHz WLAN Operation
 
Iaetsd a novel scheduling algorithms for mimo based wireless networks
Iaetsd a novel scheduling algorithms for mimo based wireless networksIaetsd a novel scheduling algorithms for mimo based wireless networks
Iaetsd a novel scheduling algorithms for mimo based wireless networks
 
03 time division-multiplexing
03 time division-multiplexing03 time division-multiplexing
03 time division-multiplexing
 
U4301106110
U4301106110U4301106110
U4301106110
 
LTE Physical Layer Transmission Mode Selection Over MIMO Scattering Channels
LTE Physical Layer Transmission Mode Selection Over MIMO Scattering ChannelsLTE Physical Layer Transmission Mode Selection Over MIMO Scattering Channels
LTE Physical Layer Transmission Mode Selection Over MIMO Scattering Channels
 

Último

Easier, Faster, and More Powerful – Alles Neu macht der Mai -Wir durchleuchte...
Easier, Faster, and More Powerful – Alles Neu macht der Mai -Wir durchleuchte...Easier, Faster, and More Powerful – Alles Neu macht der Mai -Wir durchleuchte...
Easier, Faster, and More Powerful – Alles Neu macht der Mai -Wir durchleuchte...
panagenda
 

Último (20)

Where to Learn More About FDO _ Richard at FIDO Alliance.pdf
Where to Learn More About FDO _ Richard at FIDO Alliance.pdfWhere to Learn More About FDO _ Richard at FIDO Alliance.pdf
Where to Learn More About FDO _ Richard at FIDO Alliance.pdf
 
Easier, Faster, and More Powerful – Alles Neu macht der Mai -Wir durchleuchte...
Easier, Faster, and More Powerful – Alles Neu macht der Mai -Wir durchleuchte...Easier, Faster, and More Powerful – Alles Neu macht der Mai -Wir durchleuchte...
Easier, Faster, and More Powerful – Alles Neu macht der Mai -Wir durchleuchte...
 
State of the Smart Building Startup Landscape 2024!
State of the Smart Building Startup Landscape 2024!State of the Smart Building Startup Landscape 2024!
State of the Smart Building Startup Landscape 2024!
 
Linux Foundation Edge _ Overview of FDO Software Components _ Randy at Intel.pdf
Linux Foundation Edge _ Overview of FDO Software Components _ Randy at Intel.pdfLinux Foundation Edge _ Overview of FDO Software Components _ Randy at Intel.pdf
Linux Foundation Edge _ Overview of FDO Software Components _ Randy at Intel.pdf
 
TEST BANK For, Information Technology Project Management 9th Edition Kathy Sc...
TEST BANK For, Information Technology Project Management 9th Edition Kathy Sc...TEST BANK For, Information Technology Project Management 9th Edition Kathy Sc...
TEST BANK For, Information Technology Project Management 9th Edition Kathy Sc...
 
Simplified FDO Manufacturing Flow with TPMs _ Liam at Infineon.pdf
Simplified FDO Manufacturing Flow with TPMs _ Liam at Infineon.pdfSimplified FDO Manufacturing Flow with TPMs _ Liam at Infineon.pdf
Simplified FDO Manufacturing Flow with TPMs _ Liam at Infineon.pdf
 
WSO2CONMay2024OpenSourceConferenceDebrief.pptx
WSO2CONMay2024OpenSourceConferenceDebrief.pptxWSO2CONMay2024OpenSourceConferenceDebrief.pptx
WSO2CONMay2024OpenSourceConferenceDebrief.pptx
 
FDO for Camera, Sensor and Networking Device – Commercial Solutions from VinC...
FDO for Camera, Sensor and Networking Device – Commercial Solutions from VinC...FDO for Camera, Sensor and Networking Device – Commercial Solutions from VinC...
FDO for Camera, Sensor and Networking Device – Commercial Solutions from VinC...
 
A Business-Centric Approach to Design System Strategy
A Business-Centric Approach to Design System StrategyA Business-Centric Approach to Design System Strategy
A Business-Centric Approach to Design System Strategy
 
Designing for Hardware Accessibility at Comcast
Designing for Hardware Accessibility at ComcastDesigning for Hardware Accessibility at Comcast
Designing for Hardware Accessibility at Comcast
 
WebAssembly is Key to Better LLM Performance
WebAssembly is Key to Better LLM PerformanceWebAssembly is Key to Better LLM Performance
WebAssembly is Key to Better LLM Performance
 
ERP Contender Series: Acumatica vs. Sage Intacct
ERP Contender Series: Acumatica vs. Sage IntacctERP Contender Series: Acumatica vs. Sage Intacct
ERP Contender Series: Acumatica vs. Sage Intacct
 
Continuing Bonds Through AI: A Hermeneutic Reflection on Thanabots
Continuing Bonds Through AI: A Hermeneutic Reflection on ThanabotsContinuing Bonds Through AI: A Hermeneutic Reflection on Thanabots
Continuing Bonds Through AI: A Hermeneutic Reflection on Thanabots
 
WebRTC and SIP not just audio and video @ OpenSIPS 2024
WebRTC and SIP not just audio and video @ OpenSIPS 2024WebRTC and SIP not just audio and video @ OpenSIPS 2024
WebRTC and SIP not just audio and video @ OpenSIPS 2024
 
Extensible Python: Robustness through Addition - PyCon 2024
Extensible Python: Robustness through Addition - PyCon 2024Extensible Python: Robustness through Addition - PyCon 2024
Extensible Python: Robustness through Addition - PyCon 2024
 
AI mind or machine power point presentation
AI mind or machine power point presentationAI mind or machine power point presentation
AI mind or machine power point presentation
 
Working together SRE & Platform Engineering
Working together SRE & Platform EngineeringWorking together SRE & Platform Engineering
Working together SRE & Platform Engineering
 
AI presentation and introduction - Retrieval Augmented Generation RAG 101
AI presentation and introduction - Retrieval Augmented Generation RAG 101AI presentation and introduction - Retrieval Augmented Generation RAG 101
AI presentation and introduction - Retrieval Augmented Generation RAG 101
 
Using IESVE for Room Loads Analysis - UK & Ireland
Using IESVE for Room Loads Analysis - UK & IrelandUsing IESVE for Room Loads Analysis - UK & Ireland
Using IESVE for Room Loads Analysis - UK & Ireland
 
ECS 2024 Teams Premium - Pretty Secure
ECS 2024   Teams Premium - Pretty SecureECS 2024   Teams Premium - Pretty Secure
ECS 2024 Teams Premium - Pretty Secure
 

Optimizing the Robustness of X-by-Wire using Word Combinatorics

  • 1. Optimizing the configuration of X-by-Wire networks using word combinatorics Nicolas NAVET - Bruno GAUJAL LORIA Lab. (Nancy, France) TRIO Team http://www.loria.fr/~nnavet EPFL Seminar Network Calculus Group TTP/C – Time Triggered Protocol Designed at T.U. Vienna + TTTech TTP/C main technical characteristics: • Determinism • Fault-Tolerance • Composability • Support of mode changes A good candidate for X-By-Wire .. N. NAVET - EPFL - 02/03/2004 2
  • 2. X-by-Wire Hydraulic and mechanical connection are replaced by networks and actuators Why ? Decrease of weight and cost Safety : intrusion of the steering column in the cockpit New functions : variable demultiplication - crash avoidance Less pollution (brake / transmission liquid) … N. NAVET - EPFL - 02/03/2004 3 X-by-wire : an example « Steer-by-wire » N. NAVET - EPFL - 02/03/2004 4
  • 3. A TTP/C Cluster Small Replaceable Unit (SRU) Medium Access Control : TDMA Redundant transmission support Data rate: 500kbit/s, 1Mbit/s, 2Mbit/s, 5Mbit/s, 25Mbit/s Topology: bus or star N. NAVET - EPFL - 02/03/2004 5 TDMA – Time division Multiplexed Access Slot: time window given to a station for a transmission TDMA Round: sequence of slots s.t. each station transmits exactly once Cluster Cycle: sequence of the TDMA rounds N. NAVET - EPFL - 02/03/2004 6
  • 4. TTP/C: Implications of the MAC protocol Bounded response times and « heartbeats » but: loss of bandwidth need of powerful CPU’s maximum timing contraint: If a station sends a single information, the refresh cannot be more frequent than the length of a round If a station sends several informations, the refresh cannot be more frequent than 2x the length of a round Ex: 5ms time constraint - 500kbit/s network with 200 bits per frames - at most 12 frames (6 FTUs of two nodes) or 6 frames if the station sends 2 distinct informations N. NAVET - EPFL - 02/03/2004 7 FTU: Fault Tolerant Unit FTU = set of stations that act identically FTU of cardinality 2 FTU of cardinality 3 (majority vote is possible) Replica = a frame sent by a node of the FTU N. NAVET - EPFL - 02/03/2004 8
  • 5. FTU: which protection ? Protection against: disappearance of a station (crash, disconnection..) corrupted frames (EMI) sensors or computation errors … Under the assumption of a single failure (TTP/C fault-hypothesis) : A dual redundancy ensures a protection in « the temporal domain » A triple redundancy ensures in addition a protection in « the value domain » Problem: history-state N. NAVET - EPFL - 02/03/2004 9 Goal of the study: maximize the robustness against transmission errors Transmission errors are usually highly correlated Fault Fault Tolerant Unit A Tolerant Unit B Where to place the ECU2 ECUA ECUA12 ECU2 ECUB1 replicas ? b a1 a2 b1 b2 3 a1 b1 a2 b2 b3 a1 b1 a2 b2 b3 … TTP/C Bus t TTP Round N. NAVET - EPFL - 02/03/2004 10
  • 6. Application model Ta : production cycle of the data sent by the stations of the FTU a Ta a1 a2 a3 a1 a2 a3 a1 a2 a3 TTP Round Assumptions: • no synchronization between production and transmission (round) • production cycle is a multiple of the length of a round N. NAVET - EPFL - 02/03/2004 11 Objective w.r.t. fail-silence A node is « fail-silent » if one can safely consume its data when the frame carrying the data is syntaxically correct Stations are fail-silent: « minimize Pall : the probability that all frames of the FTU sent during a production cycle will be corrupted » Stations are not fail-silent: « minimize Pone : the probability that at least one frame of the FTU will be corrupted» N. NAVET - EPFL - 02/03/2004 12
  • 7. Assumptions on the error model Each bit transmitted during an EMI will be corrupted with probability If a perturbation overlaps a whole slot, the corresponding frame is corrupted with probability 1 Starting times of the EMI bursts are independent and uniformly distributed over time The distribution of the size of the bursts is arbitrary A C B a1 a2 a3 a1 a2 a3 t TTP Round N. NAVET - EPFL - 02/03/2004 13 Objective 1 : Minimize Pone
  • 8. Majorization - Schur-Convexity vector u (u1 ,..., un ) majorizes v ( v1 ,..., vn ) if: n n k k ui vi and u[i ] v[i ] k n i 1 i 1 i 1 i 1 with (u[ i ] ,..., u[ n ] ) permutation of u s.t. u[i ] ... u[ n ] Example: (1,3,5,10) (2, 4, 4,9) n A fonction f: is Schur-convex if u v f (u ) f (v ) Schur-concave if u v f (u ) f (v ) N. NAVET - EPFL - 02/03/2004 15 Minimizing Pone I( x ) is the vector of the distance between replicas during the length of a round (sorted in ascending order) Example: I( x ) (1,1, 2) a1 a2 a3 a1 a2 a3 t TTP round Example: I( x ) (0,0, 4) a1 a2 a3 a1 a2 a3 t TTP round N. NAVET - EPFL - 02/03/2004 16
  • 9. Minimizing Pone Theorem: the best allocation for Pone is to group together all replicas (denoted allocation g) Arguments: • Pone is shur-concave: I( x ') I( x ) Pone ( x ') Pone ( x ) • I( g ) is maximum for the majorization (equal to (0,0,..., S k ) with k the number of replicas of the FTU and S the number of slots per round) N. NAVET - EPFL - 02/03/2004 17 Minimize Pone Idea of the proof (step 1): the farther the beginning of an error burst from a replica, the less likely the replica becomes corrupted. « Non-grouped » allocations have more areas close to replicas a1 a2 a3 a1 a2 a3 allocation x t a1 a2 a3 a1 a2 a3 allocation g t P( x) P( g) P( x) P( g) N. NAVET - EPFL - 02/03/2004 18
  • 10. Minimize Pone Validity of the result : • Arbitrary value and burst size distribution • Production period multiple of the round length • for all TDMA networks Combined minimization of Pone for all FTU’s is possible Robustness improvement: against a random allocation, the number of lost data is reduced from 15 to 20% on average N. NAVET - EPFL - 02/03/2004 19 Objective 2 : Minimize Pall TTP/C case
  • 11. TTP/C : the majority rule Cliques: sets of stations that disagree on the state of the network Principle: to avoid cliques, stations in the minority disconnect (« freeze ») Mechanism: before sending, a station checks that in the last round (S slots), the number of correct messages is greater than the number of incorrect messages, otherwise it disconnects ai-1 ai t S slots in a round If a station « freezes » due to transmission errors, the others follow one by one... N. NAVET - EPFL - 02/03/2004 21 TTP/C : minimize Pall Algorithm: 1) for each FTU i with Ci slots, push Ci / 2 slots in the smallest stack and Ci / 2 in the largest stack 2) concatenate the two stacks Ex: FTU A: 3 replicas – FTU B: 2 replicas – FTU C: 4 replicas stack 1 stack 2 TTP/C round : N. NAVET - EPFL - 02/03/2004 22
  • 12. TTP/C : minimize Pall Theorem: the « 2-stacks » algorithm is optimal under TTP/C Arguments: Case 1) a perturbation for each replica : identical allocation Case 2) a perturbation can corrupt several replicas with a probability decreasing in the distance between the replicas. A burst of more than S / 2 slots freezes the system, now the algorithm ensures a distance of S / 2 slots Corollary: it is useless to have more than 2 replicas per FTU if the probability to have more than one perturbation in the same round is sufficiently low N. NAVET - EPFL - 02/03/2004 23 Objective 2 : Minimize Pall TDMA case
  • 13. Balanced words A « balanced » word (or Sturmian word) is a binary sequence un n s.t. : n k m k k , n, m ui uj 1 i n j m Balanced words are computed using bracket sequences : a a un n n 1 b b where a / b is the rate of the word (nb of 1 / nb of 0) Example: balanced word of rate 3/8 (0,0,1,0,0,1,0,1) N. NAVET - EPFL - 02/03/2004 25 Multimodular functions Multimodularity [Hajek] : counterpart of convexity for discrete functions f : m Definition : Let F be a set of m+1-vectors that sum to m 0, a function f : is F-multimodular if f ( x v) f ( x w) f ( x) f ( x v w) m x et v, w F , v w Example: x (0,1,0,...,1,1,0) is a control sequence, f a cost function and v an elementary operation moving a client to the left v (0,...,0,1, 1,0,...,0) N. NAVET - EPFL - 02/03/2004 26
  • 14. Optimisation and multimodular function Global left shift operator : si ( x ) ex: s2 (0,1,0,1,1,0) (0,1,1,0,0,1) Theorem [Altman,Gaujal,Hordijk 97]: If f is multimodular m then G ( x ) 1/ m i 1 f ( si ( x )) (shift-invariant version of f ) is minimum (among all admissible sequences) if x is a balanced sequence. Theorem : If the size of the bursts is exponentially distributed then Pall is multimodular. Moreover, Pall is equal to its shift invariant version thus Pall is minimum for a balanced sequence. N. NAVET - EPFL - 02/03/2004 27 Optimal algorithm : a single FTU FTU with C replicas of size h bits in a round of total size R bits compute vi balanced word of rate C /( R C ( h 1)) x is the round initially empty If vi 1 then x : x 1...1 (h '1' concatenated) If vi 0 then x : x 0 Ex: FTU: 3 replicas of cardinality 3 in a round of size 14 vi (0,0,1,0,0,1,0,1) with rate 3/ 8 x (0,0,1,1,1,0,0,1,1,1,0,1,1,1) ( balanced word with rate 9 /14 N. NAVET - EPFL - 02/03/2004 28
  • 15. Optimal algorithm : several FTUs Problem : allocation conflicts Ex: FTU A: 3 replicas – FTU B: 2 replicas – FTU C: 1 replica x A (0,1,0,1,0,1) rate 3/ 6 xB (0,0,1,0,0,1) rate 2 / 6 xC (0,0,0,0,0,1) rate 1/ 6 An optimal allocation is still possible if • the number of cardinalities is a power of 2 • all replicas have the same size Remark: a balanced sequence is minimum for the majorization order, it is thus the worst solution for Pone! Both objectives are contradictory N. NAVET - EPFL - 02/03/2004 29 Heuristic : several FTUs The rate of emission: number of bits that FTU A must emit on average during each bit of a round : dA C Ah A / R At step i, one schedules the transmission of a frame of the FTU for which the number of due bits – number of already allocated bits is maximum N. NAVET - EPFL - 02/03/2004 30
  • 16. Pall : Heuristic vs random allocation Reduction of the number of lost messages w.r.t. a random allocation : N. NAVET - EPFL - 02/03/2004 31 Pall : Heuristic vs optimal Increase of the number of lost messages w.r.t. to the optimal : N. NAVET - EPFL - 02/03/2004 32
  • 17. Conclusion Optimal and near optimal allocation policies for TDMA and TTP/C networks Choice of the locations of the slots have a strong influence on the robustness of the network The cost function plays a major role on the shape of the solution Hypothesis on the error model are crucial Future work : Configurations made of fail-silent and non fail-silent nodes (minimizing Pone and Pall for different FTU’s) FlexRay protocol N. NAVET - EPFL - 02/03/2004 33