modeling mobile-agent-based collaborative processing in sensor networks using generalized stochastic...
TRANSCRIPT
![Page 1: Modeling Mobile-Agent-based Collaborative Processing in Sensor Networks Using Generalized Stochastic Petri Nets Hongtao Du, Hairong Qi, Gregory Peterson](https://reader036.vdocuments.mx/reader036/viewer/2022082709/56649f535503460f94c77d04/html5/thumbnails/1.jpg)
Modeling Mobile-Agent-based Collaborative Processing in SensorNetworks Using Generalized
Stochastic Petri Nets
Hongtao Du, Hairong Qi, Gregory Peterson
Department of Electrical and Computer Engineering University of Tennessee, USA
![Page 2: Modeling Mobile-Agent-based Collaborative Processing in Sensor Networks Using Generalized Stochastic Petri Nets Hongtao Du, Hairong Qi, Gregory Peterson](https://reader036.vdocuments.mx/reader036/viewer/2022082709/56649f535503460f94c77d04/html5/thumbnails/2.jpg)
Mobile-Agent-based Distributed Sensor Networks (MADSNs)
Sensors Have sensing, processing and
communication capabilities Independently sense the
environment and process data locally
Collaborate with each other to fulfill complex task
Mobile agents Dispatched from the processing
center to the sensor nodes Fuse local results during migration Perform collaborative information
processing
MADSN computing model
![Page 3: Modeling Mobile-Agent-based Collaborative Processing in Sensor Networks Using Generalized Stochastic Petri Nets Hongtao Du, Hairong Qi, Gregory Peterson](https://reader036.vdocuments.mx/reader036/viewer/2022082709/56649f535503460f94c77d04/html5/thumbnails/3.jpg)
Generalized Stochastic Petri Net (GSPN)
GSPN Advantage: modeling features of concurrency,
synchronization and randomness. Suitable for characteristics of MADSN GSPN:= (P, T, I, O, M, SI)
P: places T: transitions
I: input arc connections O: output arc connections
M: number of tokens SI: time delay of transitions
Mobile agents in distributed sensor network 1 processing element (server) and 5 sensor nodes
![Page 4: Modeling Mobile-Agent-based Collaborative Processing in Sensor Networks Using Generalized Stochastic Petri Nets Hongtao Du, Hairong Qi, Gregory Peterson](https://reader036.vdocuments.mx/reader036/viewer/2022082709/56649f535503460f94c77d04/html5/thumbnails/4.jpg)
GSPN Model for MADSN
![Page 5: Modeling Mobile-Agent-based Collaborative Processing in Sensor Networks Using Generalized Stochastic Petri Nets Hongtao Du, Hairong Qi, Gregory Peterson](https://reader036.vdocuments.mx/reader036/viewer/2022082709/56649f535503460f94c77d04/html5/thumbnails/5.jpg)
GSPN Model of Sensor Side
![Page 6: Modeling Mobile-Agent-based Collaborative Processing in Sensor Networks Using Generalized Stochastic Petri Nets Hongtao Du, Hairong Qi, Gregory Peterson](https://reader036.vdocuments.mx/reader036/viewer/2022082709/56649f535503460f94c77d04/html5/thumbnails/6.jpg)
Challenging in GSPN Modeling
Deadlock avoidance and transition selection Random selector
Our solution – ER3 transition selector Joint Entropy
Measures uncertainty of mobile agent’s migration Rolling Rocks Random Selector
Keeps fairness in transition selection
![Page 7: Modeling Mobile-Agent-based Collaborative Processing in Sensor Networks Using Generalized Stochastic Petri Nets Hongtao Du, Hairong Qi, Gregory Peterson](https://reader036.vdocuments.mx/reader036/viewer/2022082709/56649f535503460f94c77d04/html5/thumbnails/7.jpg)
Joint Entropy
Assume the probability of a mobile agent Migration success rate: 0.9, failure rate: 0.1
Joint Entropy
denotes a mobile agent migrating to the node,
Entropy rate
Gives priority to the mobile agents with higher returning probability
469.0)1(
),()()(
),,(),,(
221010
0110
p
XXXHXXHXH
XXXHXXH
p
iiii
p
iiip
iXthi nodeNp 2
469.01
469.0)1(
1
),,(limlim 0
p
p
p
XXH
p
p
p
),,( 0 pe XXHR
![Page 8: Modeling Mobile-Agent-based Collaborative Processing in Sensor Networks Using Generalized Stochastic Petri Nets Hongtao Du, Hairong Qi, Gregory Peterson](https://reader036.vdocuments.mx/reader036/viewer/2022082709/56649f535503460f94c77d04/html5/thumbnails/8.jpg)
Rolling Rocks Random (R3) Selector
Each rock (random number) has a weight between 0 and 1.
Multiple transitions conflict: multi-end seesaw
OldNewWinner RRRR )1(
(a) (b)
(c) (d)
![Page 9: Modeling Mobile-Agent-based Collaborative Processing in Sensor Networks Using Generalized Stochastic Petri Nets Hongtao Du, Hairong Qi, Gregory Peterson](https://reader036.vdocuments.mx/reader036/viewer/2022082709/56649f535503460f94c77d04/html5/thumbnails/9.jpg)
ER3 Transition Selector
: the total amount of sensor nodes
: the joint entropy
: the rock weight associated with each transition,
: the number of tokens in the input place of the transition
The transition associated with the largest will be fired.
TokensRockseNode NRRNR )(
NodeN
TokensN
RocksReR
10 RocksR
R
![Page 10: Modeling Mobile-Agent-based Collaborative Processing in Sensor Networks Using Generalized Stochastic Petri Nets Hongtao Du, Hairong Qi, Gregory Peterson](https://reader036.vdocuments.mx/reader036/viewer/2022082709/56649f535503460f94c77d04/html5/thumbnails/10.jpg)
Field Programmable Gate Array (FPGA) FPGA
Provides faster, real-time solutions
Re-configurable components at logic level 50% more time to test and verify the code 70% or more design time reduction Reduce design risk and cost For this GSPN model
3 timed and 5 immediate transition components
![Page 11: Modeling Mobile-Agent-based Collaborative Processing in Sensor Networks Using Generalized Stochastic Petri Nets Hongtao Du, Hairong Qi, Gregory Peterson](https://reader036.vdocuments.mx/reader036/viewer/2022082709/56649f535503460f94c77d04/html5/thumbnails/11.jpg)
Synthesis Procedure
Top level Configure and interconnect
re-configurable components
Register Transition Selector Conflict Controller
Structure of the top level
Design flow
![Page 12: Modeling Mobile-Agent-based Collaborative Processing in Sensor Networks Using Generalized Stochastic Petri Nets Hongtao Du, Hairong Qi, Gregory Peterson](https://reader036.vdocuments.mx/reader036/viewer/2022082709/56649f535503460f94c77d04/html5/thumbnails/12.jpg)
Conflicts Selection Comparison
First 10 transitions Overall transitions
![Page 13: Modeling Mobile-Agent-based Collaborative Processing in Sensor Networks Using Generalized Stochastic Petri Nets Hongtao Du, Hairong Qi, Gregory Peterson](https://reader036.vdocuments.mx/reader036/viewer/2022082709/56649f535503460f94c77d04/html5/thumbnails/13.jpg)
Number of Tokens at Different Time
Random selector ER3 selector
![Page 14: Modeling Mobile-Agent-based Collaborative Processing in Sensor Networks Using Generalized Stochastic Petri Nets Hongtao Du, Hairong Qi, Gregory Peterson](https://reader036.vdocuments.mx/reader036/viewer/2022082709/56649f535503460f94c77d04/html5/thumbnails/14.jpg)
Conclusions
GSPN provides a modeling tool for mobile-agent-based sensor network.
ER3 transition selector for GSPN Maximizes the modeling efficiency Balances the queue length
Synthesizing GSPN on FPGAs is a solution for complex simulations Re-configurable components improve the implementation
efficiency. More re-configurable components will be developed.