Etude de cas : Architecture multirseaux embarque dans lautomobile

1
Etude de cas Architecture multi-réseaux
embarquée dans lautomobile

• SONG YeQiong, UHP Nancy 1
• CASTELPIETRA Paolo, INPL
• SIMONOT Françoise, INPL
• LORIA-TRIO (http//www.loria.fr)

Exposé GDR - ARP - STS, Paris, 17/11/00
2
Plan

• Contexte
• Problématique
• Objectif
• Etude de Cas 1 CAN
• Etude de Cas 2 CAN et VAN
• Problèmes qui restent à traiter

3
Need of ECU in cars

• Customers wishes
• safety,
• comfort,
• cost
• Goverments requirements
• emission polution,
• fuel consumption

4
Automotive computer based control units
Airbag
AGB
Dashboard
Engine Control
Power windows
Active suspension
ABS
5
Networked control system
6
Networks inter and intra cars

• Engine Management (J1850, CAN)
• Body Electronics (comfort) (CAN, VAN)
• Entertainment (Multimedia) (MOST, IEEE1394)
• Sub bus systems (low cost CAN, LIN)
• Diagnostics (ISO 9141 and recently ISO 15765 for
  CAN)
• X-by-wire (TTP, Byteflight)
• Copper lines, fibre optic, wireless (Bluetouth)
• Vehicle-to-vehicle (wireless network)

See http//www.vector-informatik.de and
http//www.byteflight.com
7
Design method

• For software/hardware independence

Process under control
Functional architecture
Function 1
Function i
Function n
Hardware support architecture
8
Implementation architecture

• Interaction task - task and task - message

Process under control
9
Aims validation of IA

• Verification of temporal properties
• Schedulability analysis for worst case
• Simulation for any case
• Verification of dependability properties
• Reliability worst case deadline failure prob.
• Simulation
• Performance optmization
• Simulation

10
Timing constraints example

• Deadline on tasks and messages with
• HRT(Hard Real-Time) Absolute guarantee
• SRT(Soft Real-Time) Probabilistic guarantee

11
Task and message scheduling

• Local scheduling of tasks and global scheduling
  of messages

12
What we should know ?

• Message response time on network
• Task response time on processor
• End-to-end response time
• Response time when transmission errors occur
• Verification of deadline meeting gt
  Validation of application

13
Case 1 Engine management
14
Worst case message response time

• Hypothesis periodic Tx requests gt
  Peugeot-Citroen messaging

15
Worst case message response time

• Method Schedulability analysis Tin94
• Worst case message response time Rm Cm Im

where
Recurrence calculation
and
Convergence condition
16
Task modelling

• Task may transmit a message (assuming at the end
  of its execution)
• Task may wait for the reception of a message for
  starting its execution

17
Periodic task modelling

• Response time for periodic activated task

18
Message activated task model

• Response time

19
End-to-end worst case response time

• Worst case message response time Tindell et al.
  94 Rm Cm Im Jm
• Preemptive task response time

20
When Tx errors occur

• Tx errors increase the respons time
• WCDFP for measuring the robustness

(Ki max nb of retransmissions)
21
Modelling Tx error occurrence

• Error numbers during 0, t
• with
• N(t) follows a Poisson law
• u is a r.v. which follows any distribution

22
Numerical application

• P(u k) kp2(1-p)k-1
• For a 0.1 and p 0.04, Proba. of deadline
  failure

23
Periodic and aperiodic messages

• Aim minimizing the mean response time of
  aperiodic messages while guaranteeing the
  periodic message deadline meeting
• Dual-priority (dynamic priority change) is
  optimal Gaujal et Navet
• Performance

24
Conclusions on case 1

• CAN message response time
• Task response time
• End to end response time
• CAN message response time when Tx errors occur
  (combining deterministic and stochastic analysis)
• Periodic and aperiodic messages

25
Case 2 CAN and VAN
C A R O S S E
T_Engine1
T_AGB2
AGB
ABS/EPS
Suspension
WAS
CAN nw ctrl
CAN nw ctrl
CAN nw ctrl
CAN nw ctrl
CAN nw ctrl
CAN
T_ISU2
T_Y2
T_ISU3
T_Y3
CAN nw ctrl
Dashboard (ISU)
VAN nw ctrl
VAN nw ctrl
VAN nw ctrl
VAN nw ctrl
VAN
26
Message exchanges
27
Tasks and messages (part 1/2)
28
Tasks and messages (part 2/2)
29
Simulation tool
C A R O S S E
30
Causality of events
C A R O S S E
t
Node AGB
CAN network
Gateway ISU
VAN network
Node Y
End to end response time
Sequence start
Sequence end
31
Evaluation of jitters
C A R O S S E
Activation period DT
DT
DT
t
dTk1
dTk
End of instance k of T_Yi
End of instance k1 of T_Yi
End of instance k2 of T_Yi
Constraint jitters on dTk lt 50 of DT
32
Configuration of node Y
C A R O S S E
Implementation Architecture
Hardware Architecture
local IA
task running
local IA
logical task exec
executor
non-preemptive scheduler
micro processor
local IA
system object injector
task stopped
pre-emption order
task unscheduled
task ready
exec info
?P
system objects
system object manager
action
OS
Environment
I/O interface
VAN ctrl MHS 29C461
environment
I/O
node Y
CAN network
VAN network
33
Non-preemptive scheduling
C A R O S S E
Chain T_AGB2 - M11 - T_ISU2 - M13 - T_Y2
75 70 65 60 55 50 45 40 35 30 25
Termination period (ms)
18 16 14 12 10 8 6 4 2
Chain T_Engine1 - M1 - T_ISU3 - M14 - T_Y3
34
Modification of scheduler
C A R O S S E
Implementation Architecture
Hardware Architecture
local IA
task running
local IA
logical task exec
executor
non-preemptive scheduler
micro processor
preemptive scheduler (OSEK/VDX)
local IA
system object injector
task stopped
pre-emption order
task unscheduled
task ready
exec info
?P
system objects
system object manager
action
OS
Environment
I/O interface
VAN ctrl MHS 29C461
environment
I/O
node Y
CAN network
VAN network
35
Preemptive scheduling
C A R O S S E
Chain T_AGB2 - M11 - T_ISU2 - M13 - T_Y2
75 70 65 60 55 50 45 40 35 30 25
Termination period (ms)
18 16 14 12 10 8 6 4 2
Chain T_Engine1 - M1 - T_ISU3 - M14 - T_Y3
36
Conclusions on Case 2
C A R O S S E

• A modular approach to model the implementation
  architecture
• A methodology to easily build and use models
• A library of modules
• CAN network and controllers
• VAN network and controllers
• TTP/C network and controllers
• OSEK scheduler
• etc.

37
What else to do ?

• Worst case schedulability analysis of the case 2
• Optimal task and message priority assignement
  (for minimizing aperiodic message response time
  or for maximizing the robustness when Tx errors
  occur)
• Other scheduling policies to reduce the jitters
• Finding a good compromising between periodic and
  event activation of tasks (PhD Thesis of F.
  Jumel)

38
Future trends

• X-by-wire
• Drive-by-wire systems (such as brake- and
  steer-by-wire) is replacing hydraulic and
  mechanical linkages for steering, brakes,
  throttle and suspension
• need for time-critical and safety-critical
  networks (candidates Byteflight, TTP)
• Gateways for interconnecting the different
  in-vehicle networks but also with Internet
• FastEthernet, ATM and Bluetooth candidates for
  in-vehicle networks ?