modeling multi-periodic simulink systems by synchronous...
TRANSCRIPT
![Page 1: Modeling multi-periodic Simulink systems by Synchronous ...2016.rtas.org/wp-content/uploads/2016/06/26.pdf · Modeling multi-periodic Simulink systems by Synchronous Dataflow Graphs](https://reader034.vdocument.in/reader034/viewer/2022051601/5acffe887f8b9ae2138d3436/html5/thumbnails/1.jpg)
Modeling multi-periodic Simulink systems by Synchronous Dataflow GraphsEnagnon C. Klikpo (IRT SystemX)
Jad Kathib (CEA)
Alix Munier-Kordon (UPMC)
![Page 2: Modeling multi-periodic Simulink systems by Synchronous ...2016.rtas.org/wp-content/uploads/2016/06/26.pdf · Modeling multi-periodic Simulink systems by Synchronous Dataflow Graphs](https://reader034.vdocument.in/reader034/viewer/2022051601/5acffe887f8b9ae2138d3436/html5/thumbnails/2.jpg)
CONFIDENTIEL
Context
Advanced embedded technologies in modern cars Standards (emission, safety),
ADAS, connected or autonomous car
Need of computing power
AUTomotive Open System Architecture : Standard for the design and development of automotive E/E architecture
AUTOSAR 4.x introduced multicore platforms
Multicore for critical automotive application raises some issues The mastery of the dataflow (and timing) among functionalities over cores
Missing a dataflow can lead to fatal scenario: e.g crash detection and inflator (ACU)
Need of computing power and multicore issues
2
![Page 3: Modeling multi-periodic Simulink systems by Synchronous ...2016.rtas.org/wp-content/uploads/2016/06/26.pdf · Modeling multi-periodic Simulink systems by Synchronous Dataflow Graphs](https://reader034.vdocument.in/reader034/viewer/2022051601/5acffe887f8b9ae2138d3436/html5/thumbnails/3.jpg)
CONFIDENTIEL
Mastering of the dataflow
Simulink models Synchronous sequential execution
dataflow communication patterns
Need of predictability
Understand and model the communication Synchronous dataflow graph (SDFG)
Need of a dataflow formalism
3
![Page 4: Modeling multi-periodic Simulink systems by Synchronous ...2016.rtas.org/wp-content/uploads/2016/06/26.pdf · Modeling multi-periodic Simulink systems by Synchronous Dataflow Graphs](https://reader034.vdocument.in/reader034/viewer/2022051601/5acffe887f8b9ae2138d3436/html5/thumbnails/4.jpg)
CONFIDENTIEL
Plan
Reminder on SDFG
Description of communication in Simulink
Identification of Simulink communication patterns
Correspondence between Simulink and SDFG
Example of a Fuel Cell Control System
Conclusion
Modeling multi-periodic Simulink systems by Synchronous Dataflow Graphs
4
![Page 5: Modeling multi-periodic Simulink systems by Synchronous ...2016.rtas.org/wp-content/uploads/2016/06/26.pdf · Modeling multi-periodic Simulink systems by Synchronous Dataflow Graphs](https://reader034.vdocument.in/reader034/viewer/2022051601/5acffe887f8b9ae2138d3436/html5/thumbnails/5.jpg)
CONFIDENTIEL
The SDFG formalism
SDFG: Directed graph Lee et Messerschmidtt
Modeling communications in data flow applications
Static description: Each process has the same behavior during execution Low expressivity
Completely predictive
Overview
5
![Page 6: Modeling multi-periodic Simulink systems by Synchronous ...2016.rtas.org/wp-content/uploads/2016/06/26.pdf · Modeling multi-periodic Simulink systems by Synchronous Dataflow Graphs](https://reader034.vdocument.in/reader034/viewer/2022051601/5acffe887f8b9ae2138d3436/html5/thumbnails/6.jpg)
CONFIDENTIEL
Simulink
Matlab/Simulink: specification and simulation tool used in Industry
A Simulink system is a set of communicating blocks
Blocks are executed at they sample time (their period)
B A
DCInput port
Output port
5ms 3ms
8ms 4ms
6
![Page 7: Modeling multi-periodic Simulink systems by Synchronous ...2016.rtas.org/wp-content/uploads/2016/06/26.pdf · Modeling multi-periodic Simulink systems by Synchronous Dataflow Graphs](https://reader034.vdocument.in/reader034/viewer/2022051601/5acffe887f8b9ae2138d3436/html5/thumbnails/7.jpg)
CONFIDENTIEL
Simulink
Block execution consists in: Input update and outputs computation (depend on the state and/or the inputs)
Updating the block state
Several communication mechanisms in Simulink: The order in which blocks are executed
The input data that each execution of a block uses
We have extracted three main communication patterns « Direct » communication
« Delayed » communication
« Hybrid » communication
Communication patterns
7
![Page 8: Modeling multi-periodic Simulink systems by Synchronous ...2016.rtas.org/wp-content/uploads/2016/06/26.pdf · Modeling multi-periodic Simulink systems by Synchronous Dataflow Graphs](https://reader034.vdocument.in/reader034/viewer/2022051601/5acffe887f8b9ae2138d3436/html5/thumbnails/8.jpg)
CONFIDENTIEL
SDFG model of Simulink multi-periodic systemsHybrid communication
0ms
B2 A4A1 A2 A3 A5B1 B3 A6B4
3ms 5ms 6ms 9ms 10ms 12ms 15ms
A B
3ms 5ms
Hybrid 3 5
A B53
8
A
B
5
5
8
10 15 20
11 14 17 20 23
B1 B2 B3 B4
A1 A2 A3 A4 A5 A6
![Page 9: Modeling multi-periodic Simulink systems by Synchronous ...2016.rtas.org/wp-content/uploads/2016/06/26.pdf · Modeling multi-periodic Simulink systems by Synchronous Dataflow Graphs](https://reader034.vdocument.in/reader034/viewer/2022051601/5acffe887f8b9ae2138d3436/html5/thumbnails/9.jpg)
CONFIDENTIEL
Fuel Cell Control System
9
![Page 10: Modeling multi-periodic Simulink systems by Synchronous ...2016.rtas.org/wp-content/uploads/2016/06/26.pdf · Modeling multi-periodic Simulink systems by Synchronous Dataflow Graphs](https://reader034.vdocument.in/reader034/viewer/2022051601/5acffe887f8b9ae2138d3436/html5/thumbnails/10.jpg)
CONFIDENTIEL
Conclusion and perspectives
Formal equivalence between Simulink and SDFG SDFG results for Simulink systems implementation
SDFG is widely used: Initially design to for dataflow application (signal processing)
Compilation on multi-core (with several variants: CSDF, HSDF) Special case of petri nets (basic)
It has proven effective for modeling application flow
SDFG has existing results on Scheduling and mapping
Resources optimization
10
![Page 11: Modeling multi-periodic Simulink systems by Synchronous ...2016.rtas.org/wp-content/uploads/2016/06/26.pdf · Modeling multi-periodic Simulink systems by Synchronous Dataflow Graphs](https://reader034.vdocument.in/reader034/viewer/2022051601/5acffe887f8b9ae2138d3436/html5/thumbnails/11.jpg)
CONFIDENTIEL
Conclusion and perspectives
SDFG rather than Simulink models
Preemptive Real-time implementation Use of mathematical tools of SDF
Other approaches and constraints Language bases approaches
PRELUDE
We are constrained by Simulink
Perspectives
11
Requirements:- Hardware- Safety
AUTOSAR configurations
Arxml, source codes, …
Analyze & Implementation
SDFG
![Page 12: Modeling multi-periodic Simulink systems by Synchronous ...2016.rtas.org/wp-content/uploads/2016/06/26.pdf · Modeling multi-periodic Simulink systems by Synchronous Dataflow Graphs](https://reader034.vdocument.in/reader034/viewer/2022051601/5acffe887f8b9ae2138d3436/html5/thumbnails/12.jpg)
Modeling multi-periodic Simulink systems by Synchronous Dataflow GraphsEnagnon C. Klikpo
![Page 13: Modeling multi-periodic Simulink systems by Synchronous ...2016.rtas.org/wp-content/uploads/2016/06/26.pdf · Modeling multi-periodic Simulink systems by Synchronous Dataflow Graphs](https://reader034.vdocument.in/reader034/viewer/2022051601/5acffe887f8b9ae2138d3436/html5/thumbnails/13.jpg)
CONFIDENTIEL
AnnexTransformation principle
Modeling principle (equivalence) precedence constraints
data dependencies
The obtained data dependencies equation
𝜔 : periods
𝑔𝑐𝑑 : greatest common divisor
13
![Page 14: Modeling multi-periodic Simulink systems by Synchronous ...2016.rtas.org/wp-content/uploads/2016/06/26.pdf · Modeling multi-periodic Simulink systems by Synchronous Dataflow Graphs](https://reader034.vdocument.in/reader034/viewer/2022051601/5acffe887f8b9ae2138d3436/html5/thumbnails/14.jpg)
CONFIDENTIEL
Direct communication
Delayed communication
Hybrid communication
B1
AnnexSimulink communication mechanisms
A B
3ms 5ms
DirectB2 A4A1 A2 A3 A5B1 B3 A6 B4
0ms 3ms 5ms 6ms 9ms 10ms 12ms 15ms
A B
3ms 5ms
B2 A4A1 A2 A3 A5B3 DelayedA6B4
0ms 3ms 5ms 6ms 9ms 10ms 12ms 15ms
A B
3ms 5ms
HybridB2 A4A1 A2 A3 A5B1 B3 A6B4
0ms 3ms 5ms 6ms 9ms 10ms 12ms 15ms
14