Cyber-Physical Systems and Mixed Simulations Tran Van Hoang - - PowerPoint PPT Presentation
Cyber-Physical Systems and Mixed Simulations Tran Van Hoang Supervisor: Professor Bernard Pottier University of Western Brittany June 23, 2015 Tran Van Hoang (UBO) June 23, 2015 1 / 23 Contents Simulating and monitoring the environment 1
Tran Van Hoang
Supervisor: Professor Bernard Pottier University of Western Brittany
June 23, 2015
Tran Van Hoang (UBO) June 23, 2015 1 / 23
1
Simulating and monitoring the environment
2
Problems of scheduling simulations of physical and sensing systems
3
Synthesizing cell system
4
Parallel simulations on the GPU
5
Mixed simulations with HLA
6
Conclusions
Tran Van Hoang (UBO) June 23, 2015 2 / 23
Climate changes increasingly becomes critical. ֒ → These changes have to be observed and predicted. Wireless Sensor Networks (WSNs) have recently emerged in monitoring the environment. ֒ → Such systems also need to be validated before implementations. Simulating and designing WSN for accurate observation of the environment.
Tran Van Hoang (UBO) June 23, 2015 3 / 23
The lack of interoperability between WSN simulations and physical simulations. Modeling and simulating physical systems are difficult and time-consuming because these systems have:
◮ a large size. ◮ complicated behaviours. Tran Van Hoang (UBO) June 23, 2015 4 / 23
Figure 1: A region taken from OpenStreetMap.
To employ the GPU capability. To mix and schedule simulations with HLA (High Level Architecture).
Tran Van Hoang (UBO) June 23, 2015 5 / 23
PickCell tool enables to generate cell networks as a set of processes and neighbourhood connections. The number of neighbours depend on defined Cellular Automata pattern.
Figure 2: A cell network with Von Neumann connection.
Tran Van Hoang (UBO) June 23, 2015 6 / 23
A method based on PickCell tool.
Figure 3: A process of modeling physical systems.
Tran Van Hoang (UBO) June 23, 2015 7 / 23
GPU contains hundreds of cores running in parallel. It is good for many similar computations.
Figure 4: Comparison between CPU and GPU.
Tran Van Hoang (UBO) June 23, 2015 8 / 23
For the physical and network simulations, cells must compute state at time t+1 according to:
◮ local state. ◮ neighbours’ states. ◮ transition function.
The computation is moved on the GPU.
Tran Van Hoang (UBO) June 23, 2015 9 / 23
Figure 5: Fire expansion simulation.
Tran Van Hoang (UBO) June 23, 2015 10 / 23
A simulation of fire expansion.
1
Using PickCell to extract a cell network from the map.
2
Cell state: [TREE, FIRE, ASH, or EMPTY]
3
Transition function: computed on the GPU.
Kernel execution
if (nowState == TREE && firedNeighbour() == true) nextState = FIRE; if (nowState == FIRE) nextState = ASH; if (nowState == ASH) nextState = EMPTY; return nextState; Tran Van Hoang (UBO) June 23, 2015 11 / 23
Hardware
◮ CPU: Intel(R) Xeon(R), 3.40GHz, 8CPU x 4cores. ◮ GPU: GeForce GTX 680, 1536 cores, 1024 threads.
Physical model used: Pollution diffusion in the river.
Figure 6: Pollution diffusion simulation.
Tran Van Hoang (UBO) June 23, 2015 12 / 23
Result:
Figure 7: Comparison of time execution between CPU and GPU.
Tran Van Hoang (UBO) June 23, 2015 13 / 23
High Level Architecture (HLA) is an IEEE standard for distributed simulations. Cell system and WSN simulations can communicate together.
Figure 8: HLA architecture.
Tran Van Hoang (UBO) June 23, 2015 14 / 23
HLA defines three main components: Rules.
◮ A set of rules which must be followed to achieve proper interaction of
simulations in a federation.
Interface specification.
◮ To prescribes the interface functions between RTI and the simulations..
Object Model Template (OMT).
◮ To defines the way federations and federates have to be documented. Tran Van Hoang (UBO) June 23, 2015 15 / 23
A typical Federation Execution.
1 Initialize federation. ◮ Create Federation. ◮ Join Federation. 2 Declare objects. ◮ Publish Object Class Attributes. ◮ Subscribe Object Class Attributes. 3 Loop execution and exchange information. ◮ Update/Reflect Attributes Values. ◮ Time Advance Request, Time Advance Grant. 4 Terminate execution. ◮ Resign Federation. ◮ Destroy Federation. Tran Van Hoang (UBO) June 23, 2015 16 / 23
Figure 9: A proposed federation.
Tran Van Hoang (UBO) June 23, 2015 17 / 23
Federate Time constrained Time regulating Time advance Forest Yes Yes Time stepped WSN Yes Yes Time stepped Visualization Yes Yes/No Time stepped
Table 1: Time management of the federation.
Object Class Attributes Published by Subscribed by ForestNode State, Position ForestNode WSN, Visualization WSNNode State, Position WSNNode Visualization
Table 2: Objects and their attributes, publishers and subscribers.
Tran Van Hoang (UBO) June 23, 2015 18 / 23
How the federates exchange information together?
Figure 10: A communication between Forest federate and WSN federate.
Tran Van Hoang (UBO) June 23, 2015 19 / 23
Figure 11: Three simulations: Forest fire, WSN, and Visualization.
Tran Van Hoang (UBO) June 23, 2015 20 / 23
Conclusions
◮ Synthesizing cell systems based on PickCell tool to develop parallel
simulations.
◮ Parallel simulations concurrently run together. This can lead to
accurate observation of the environment.
Future works
◮ The proposed method will be applied in a real simulation project (Mad
Environment company).
◮ The multiple dimension data will also be considered in future work. Tran Van Hoang (UBO) June 23, 2015 21 / 23
Teodora Sanislav, Liviu Miclea. Cyber-Physical Systems - Concept, Challenges and Research Areas. CEAI Vol.14, No.2, pp. 28-33, 2012. Bernard Pottier, Pierre-Yves Lucas. Dynamic networks NetGen: objectives, installation, use, and programming. Universit´ e de Bretagne Occidentale. August 26, 2014.
Introduction to the High Level Architecture. Prentice Hall, 1999. The study region, Mekong Delta Region, South of Vietnam. https://www.openstreetmap.org NVIDIA, Graphics Processing Unit (GPU). http://www.nvidia.com/object/gpu.html. Common-pool Resources and Multi-Agent Simulations (CORMAS) CIRAD research
GeForce GTX 680. http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-680/specifications.
Tran Van Hoang (UBO) June 23, 2015 22 / 23
Tran Van Hoang (UBO) June 23, 2015 23 / 23