SlideShare una empresa de Scribd logo
1 de 18
Distributed target tracking
under realistic network conditions


 Christian Nastasi                     Andrea Cavallaro
  c.nastasi@sssup.it              andrea.cavallaro@eecs.qmul.ac.uk




        www.eecs.qmul.ac.uk/~andrea/wise-mnet.html
Context

 • Wireless Sensor Networks (WSNs)
        –   networks of autonomous sensors used for pervasive applications
        –   large-number deployments, highly scalable
        –   resource-constrained
        –   scalar data (e.g. temperature, light, pressure)


 The way ahead...

 • Wireless Multimedia Sensor Networks (WMSNs)
        – vectorial data (e.g. audio, video)
        – raw data cannot (always) be transferred
        – local processing is required (but much more complex!)




C. Nastasi, A. Cavallaro                        Distributed target tracking under realistic network conditions
Application: multi-sensor tracking

 • Objective
               Continuous estimation of the target state
               given a set of measurements (observations)
               obtained from spatially distributed sensing nodes.


                                                                                                   4
                                                                                                  zk
 Z k  ( zk
          1                 2    N
                           zk  zk )               1
                                                  zk
Measurements


                                                2
                                               zk
  xk  f(Z k , Z1:k 1, x0:k-1 )
                                                                                              3
 State estimation                                                                            zk


C. Nastasi, A. Cavallaro                       Distributed target tracking under realistic network conditions
Approaches




              Centralized                           Decentralized                           Distributed


   Distributed and decentralized multi-camera tracking
   M. Taj, A. Cavallaro
   IEEE Signal Processing Magazine, Vol. 28, Issue 3, May 2011




C. Nastasi, A. Cavallaro                                         Distributed target tracking under realistic network conditions
Distributed tracking: strategies

 • Distributed target tracking
        – need a collaborative information exchange mechanism
        – consensus-based algorithms
               • Parallel (e.g. Kalman Consensus Filter [Olfati-Saber2005], Distributed Particle
                 Filters [Gu2007])
        – data aggregation algorithms
               • Sequential (e.g. Distributed Particle Filters [Hlinka2009])
                                                                           estimate

                                                        start




                      consensus                                              aggregation
C. Nastasi, A. Cavallaro                                 Distributed target tracking under realistic network conditions
Distributed Particle Filters (DPFs)

 •    Basic ideas:
        – each node executes a local Particle Filter (PF)
        – measurements are synchronized, calibration is known
        – some information is exchanged

 •    Likelihood sharing [Coates2004]
        – exchange information to have a common model of the likelihood
        – random number generators are synchronized

 •    Posterior sharing
        – the network has a common knowledge of the posterior pdf
        – consensus-based approach [Sheng2005, Gu2007]
        – aggregation-based approach [Sheng2005, Hlinka2009]
               • spatial sequence of aggregation steps
               • Partial Posterior (PP) is exchanged among the nodes


C. Nastasi, A. Cavallaro                             Distributed target tracking under realistic network conditions
Aggregation-based DPF

                                                            f(x k | Z1:k 1 , z k:4 )  f(x k | Z k )
                                                                                1


                         1
     f(x k | Z1:k 1 , z k )        start          estimate
                                                                       4
                                                                      zk
                                       1
                                      zk


                                2
                               zk
   f(x k | Z1:k 1 , z k:2 )
                       1

                                                               3
                                                              zk
                                                                            f(x k | Z1:k 1 , z k:3 )
                                                                                                1




                Problem: Particle dissemination is not feasible!
                Solution: Gaussian Mixture Model of the Partial Posterior (GMM-PP)
                          Independence from the # of particles

C. Nastasi, A. Cavallaro                         Distributed target tracking under realistic network conditions
How to extend this tracking approach
      from WSNs to WMSNs?
Proposed approach

   • Objective
          – Distributed tracking under realistic conditions in camera-based WMSNs


   • Problems
          – existing approaches are theoretical and designed for WSNs
          – need adaptation for limited Field-Of-View sensors (cameras)
                 • detection miss
                 • target hand-over
                 • target loss
          – need mechanisms for the definition of the aggregation chain
                 • first node (starts iteration)
                 • intermediate nodes (aggregate local measurement to the PP)
                 • last node (performs estimation)
          – a network-simulator environment is required



C. Nastasi, A. Cavallaro                             Distributed target tracking under realistic network conditions
First node

                                x          (i)
                                         k -1     ,w              
                                                              (i) P
                                                              k -1 i 1
                                                                                       1. Knows previous posterior
                                     1
                                z    k                                                    and local measurement

                                                                                       2. Prediction and Update:
                                                                                             •   re-sampling
                                                                                             •   draw from state-transition
                                                                                             •   weight update from likelihood

  next-hop                                                                             3. GMM-PP creation


             x                               
                          (i)        (i) P
                                                                                       4. Next-hop selection
                        k -1,w       k -1 i 1
                                                                                       5. Sends GMM-PP
              x   (i)
                  k      ~ f(x        k   |x      (i)
                                                  k -1    )     i  1,..., P

                                          1
                                    f(z k|x k )
                                                  (i)                            x   (i)
                                                                                      k     ,w   
                                                                                             (i) P
                                                                                             k i 1
                                                                                                                    f   1
                                                                                                                        GMM - PP
             w (i)
               k               P             1         (j)
                                                                 i  1,..., P
                                 f(z k|x k )
C. Nastasi, A. Cavallaroj 1                                                     Distributed target tracking under realistic network conditions
Intermediate node h

next-hop
                                                                               1. Receives PP from node h-1

                                                                               2. Importance sampling:
                                                                                      •        use the incoming PP as
                                                                                               importance function g()
                                                                                      •        draw from importance function
                                                                                      •        weight update: CONDENSATION

                            h
                           zk                                                  3. GMM-PP creation

                                                                               4. Next-hop selection
     g(x k )  f     1:h -1
                     GMM - PP     (x k | z            1:h -1
                                                      k        )
                                                                               5. Sends GMM-PP
     x (i)
       k     ~f   GMM - PP (x k | z k:h -1 )
                                    1
                                                               i  1,..., P


             w (i)
               k          P
                                  h
                               f(z k|x k )
                                          (i)

                                                               i  1,..., P
                                                                                x       (i)
                                                                                      k -1     ,w     
                                                                                                  (i) P
                                                                                                  k -1 i 1
                                                                                                                         1:h
                                                                                                                       f GMM - PP
                                      h         (j)
                            f(z k|x k )
                           j 1
C. Nastasi, A. Cavallaro                                                       Distributed target tracking under realistic network conditions
Last node
First node
  at k+1
                                                1. Receives PP from node N-1

                                                2. Importance sampling as for
 N
zk                                                 intermediate nodes

                                                3. Last PP is also the global PP

                                                4. Target state estimation

                                                5. Next tracking step starts here!


  After importance sampling:   f PPN  f(x k | Z k )
                                 1:

                                     P
  Estimation:                  x k   wk x k
                               ˆ        (i)  (i)

                                     i 1



C. Nastasi, A. Cavallaro                    Distributed target tracking under realistic network conditions
Experimental setups

 • Simulations
        •    number of nodes: N = 10, 50, 100, 300, 500, 700, 1000
        •    number of particles: P = 100, 300, 500

        •    DPF with different GMM configurations
              •    No GMM approximation: DPF-0
              •    Variable number of GMM components: DPF-1, DPF-5


        •    realistic network conditions



         Simulator: WiSE-MNet        www.eecs.qmul.ac.uk/~andrea/wise-mnet.html



C. Nastasi, A. Cavallaro                         Distributed target tracking under realistic network conditions
Simulation setup

 • Network
        –   T-MAC protocol, BW = 250 kbps
        –   request-to-send/clear-to-send mechanism
        –   acknowledged-transmission mechanism
        –   number of retransmissions: 10


 • Cameras
        – Covering 6000 sqm (random uniform distribution)
        – Top-down facing cameras: 6m from the ground plane (FOV is 10m X 6m)
        – Frame rate = 1fps


 • 100 simulation runs, each of 10 minutes


C. Nastasi, A. Cavallaro                   Distributed target tracking under realistic network conditions
What do we measure?

                              K tr       Ktr # of estimations (detected events)
• Estimation efficiency    E
                              K          K # of observations (all the events)


                                              K tr
• Average estimation delay D  1              d(k)
                              K tr            i 1

                           d(k) :   Estimation delay for the k-th tracking step




C. Nastasi, A. Cavallaro              Distributed target tracking under realistic network conditions
Efficiency
Network Delay
Conclusions

 • Conclusions
        – distributed target tracking for camera-based WMSNs with a DPF
               • Dealing with limited-FOV sensors
               • Operating on a network-simulator environment
        – importance of co-design between tracking algorithms and
          communication protocols

                          Simulator available as open source at
                       www.eecs.qmul.ac.uk/~andrea/wise-mnet.html


 • Future work
        – Comparing other state-of-the-art protocols (e.g. consensus-based)
        – Using the full vision-pipeline: more complex features



C. Nastasi, A. Cavallaro                          Distributed target tracking under realistic network conditions

Más contenido relacionado

La actualidad más candente

150424 Scalable Object Detection using Deep Neural Networks
150424 Scalable Object Detection using Deep Neural Networks150424 Scalable Object Detection using Deep Neural Networks
150424 Scalable Object Detection using Deep Neural Networks
Junho Cho
 
"Semantic Segmentation for Scene Understanding: Algorithms and Implementation...
"Semantic Segmentation for Scene Understanding: Algorithms and Implementation..."Semantic Segmentation for Scene Understanding: Algorithms and Implementation...
"Semantic Segmentation for Scene Understanding: Algorithms and Implementation...
Edge AI and Vision Alliance
 

La actualidad más candente (20)

#6 PyData Warsaw: Deep learning for image segmentation
#6 PyData Warsaw: Deep learning for image segmentation#6 PyData Warsaw: Deep learning for image segmentation
#6 PyData Warsaw: Deep learning for image segmentation
 
Pr057 mask rcnn
Pr057 mask rcnnPr057 mask rcnn
Pr057 mask rcnn
 
Intepretability / Explainable AI for Deep Neural Networks
Intepretability / Explainable AI for Deep Neural NetworksIntepretability / Explainable AI for Deep Neural Networks
Intepretability / Explainable AI for Deep Neural Networks
 
ISCAS'18: A Deep Neural Network on the Nested RNS (NRNS) on an FPGA: Applied ...
ISCAS'18: A Deep Neural Network on the Nested RNS (NRNS) on an FPGA: Applied ...ISCAS'18: A Deep Neural Network on the Nested RNS (NRNS) on an FPGA: Applied ...
ISCAS'18: A Deep Neural Network on the Nested RNS (NRNS) on an FPGA: Applied ...
 
Final Poster
Final PosterFinal Poster
Final Poster
 
Deep learning for object detection
Deep learning for object detectionDeep learning for object detection
Deep learning for object detection
 
DeepLab V3+: Encoder-Decoder with Atrous Separable Convolution for Semantic I...
DeepLab V3+: Encoder-Decoder with Atrous Separable Convolution for Semantic I...DeepLab V3+: Encoder-Decoder with Atrous Separable Convolution for Semantic I...
DeepLab V3+: Encoder-Decoder with Atrous Separable Convolution for Semantic I...
 
Convolutional neural networks 이론과 응용
Convolutional neural networks 이론과 응용Convolutional neural networks 이론과 응용
Convolutional neural networks 이론과 응용
 
150424 Scalable Object Detection using Deep Neural Networks
150424 Scalable Object Detection using Deep Neural Networks150424 Scalable Object Detection using Deep Neural Networks
150424 Scalable Object Detection using Deep Neural Networks
 
Learning visual representation without human label
Learning visual representation without human labelLearning visual representation without human label
Learning visual representation without human label
 
Image Retrieval (D4L5 2017 UPC Deep Learning for Computer Vision)
Image Retrieval (D4L5 2017 UPC Deep Learning for Computer Vision)Image Retrieval (D4L5 2017 UPC Deep Learning for Computer Vision)
Image Retrieval (D4L5 2017 UPC Deep Learning for Computer Vision)
 
Batch normalization presentation
Batch normalization presentationBatch normalization presentation
Batch normalization presentation
 
FPGA2018: A Lightweight YOLOv2: A binarized CNN with a parallel support vecto...
FPGA2018: A Lightweight YOLOv2: A binarized CNN with a parallel support vecto...FPGA2018: A Lightweight YOLOv2: A binarized CNN with a parallel support vecto...
FPGA2018: A Lightweight YOLOv2: A binarized CNN with a parallel support vecto...
 
Vision and Multimedia Reading Group: DeCAF: a Deep Convolutional Activation F...
Vision and Multimedia Reading Group: DeCAF: a Deep Convolutional Activation F...Vision and Multimedia Reading Group: DeCAF: a Deep Convolutional Activation F...
Vision and Multimedia Reading Group: DeCAF: a Deep Convolutional Activation F...
 
Review-image-segmentation-by-deep-learning
Review-image-segmentation-by-deep-learningReview-image-segmentation-by-deep-learning
Review-image-segmentation-by-deep-learning
 
Lecture 7: Recurrent Neural Networks
Lecture 7: Recurrent Neural NetworksLecture 7: Recurrent Neural Networks
Lecture 7: Recurrent Neural Networks
 
Deep Learning for Computer Vision: Image Retrieval (UPC 2016)
Deep Learning for Computer Vision: Image Retrieval (UPC 2016)Deep Learning for Computer Vision: Image Retrieval (UPC 2016)
Deep Learning for Computer Vision: Image Retrieval (UPC 2016)
 
"Semantic Segmentation for Scene Understanding: Algorithms and Implementation...
"Semantic Segmentation for Scene Understanding: Algorithms and Implementation..."Semantic Segmentation for Scene Understanding: Algorithms and Implementation...
"Semantic Segmentation for Scene Understanding: Algorithms and Implementation...
 
(Paper Review) Reconstruction of Monte Carlo Image Sequences using a Recurren...
(Paper Review) Reconstruction of Monte Carlo Image Sequences using a Recurren...(Paper Review) Reconstruction of Monte Carlo Image Sequences using a Recurren...
(Paper Review) Reconstruction of Monte Carlo Image Sequences using a Recurren...
 
Deep LearningフレームワークChainerと最近の技術動向
Deep LearningフレームワークChainerと最近の技術動向Deep LearningフレームワークChainerと最近の技術動向
Deep LearningフレームワークChainerと最近の技術動向
 

Similar a Distributed target tracking under realistic network conditions

FR1.L09 - PREDICTIVE QUANTIZATION OF DECHIRPED SPOTLIGHT-MODE SAR RAW DATA IN...
FR1.L09 -	PREDICTIVE QUANTIZATION OF DECHIRPED SPOTLIGHT-MODE SAR RAW DATA IN...FR1.L09 -	PREDICTIVE QUANTIZATION OF DECHIRPED SPOTLIGHT-MODE SAR RAW DATA IN...
FR1.L09 - PREDICTIVE QUANTIZATION OF DECHIRPED SPOTLIGHT-MODE SAR RAW DATA IN...
grssieee
 
Deep Learning
Deep LearningDeep Learning
Deep Learning
Pierre de Lacaze
 
IOEfficientParalleMatrixMultiplication_present
IOEfficientParalleMatrixMultiplication_presentIOEfficientParalleMatrixMultiplication_present
IOEfficientParalleMatrixMultiplication_present
Shubham Joshi
 

Similar a Distributed target tracking under realistic network conditions (20)

Surveys
SurveysSurveys
Surveys
 
UCB 2012-02-28
UCB 2012-02-28UCB 2012-02-28
UCB 2012-02-28
 
FR1.L09 - PREDICTIVE QUANTIZATION OF DECHIRPED SPOTLIGHT-MODE SAR RAW DATA IN...
FR1.L09 -	PREDICTIVE QUANTIZATION OF DECHIRPED SPOTLIGHT-MODE SAR RAW DATA IN...FR1.L09 -	PREDICTIVE QUANTIZATION OF DECHIRPED SPOTLIGHT-MODE SAR RAW DATA IN...
FR1.L09 - PREDICTIVE QUANTIZATION OF DECHIRPED SPOTLIGHT-MODE SAR RAW DATA IN...
 
Target tracking suing multiple auxiliary particle filtering
Target tracking suing multiple auxiliary particle filteringTarget tracking suing multiple auxiliary particle filtering
Target tracking suing multiple auxiliary particle filtering
 
Deep Learning
Deep LearningDeep Learning
Deep Learning
 
Convolutional Neural Networks - Veronica Vilaplana - UPC Barcelona 2018
Convolutional Neural Networks - Veronica Vilaplana - UPC Barcelona 2018Convolutional Neural Networks - Veronica Vilaplana - UPC Barcelona 2018
Convolutional Neural Networks - Veronica Vilaplana - UPC Barcelona 2018
 
Wqtc2011 causes offalsealarms-20111115-final
Wqtc2011 causes offalsealarms-20111115-finalWqtc2011 causes offalsealarms-20111115-final
Wqtc2011 causes offalsealarms-20111115-final
 
Massively Parallel K-Nearest Neighbor Computation on Distributed Architectures
Massively Parallel K-Nearest Neighbor Computation on Distributed Architectures Massively Parallel K-Nearest Neighbor Computation on Distributed Architectures
Massively Parallel K-Nearest Neighbor Computation on Distributed Architectures
 
UC Berkeley 4.19
UC Berkeley 4.19UC Berkeley 4.19
UC Berkeley 4.19
 
lecture5.pdf
lecture5.pdflecture5.pdf
lecture5.pdf
 
PanicO
PanicOPanicO
PanicO
 
Implementation of the fully adaptive radar framework: Practical limitations
Implementation of the fully adaptive radar framework: Practical limitationsImplementation of the fully adaptive radar framework: Practical limitations
Implementation of the fully adaptive radar framework: Practical limitations
 
Conditional Image Generation with PixelCNN Decoders
Conditional Image Generation with PixelCNN DecodersConditional Image Generation with PixelCNN Decoders
Conditional Image Generation with PixelCNN Decoders
 
Ad hoc routing
Ad hoc routingAd hoc routing
Ad hoc routing
 
モデル高速化百選
モデル高速化百選モデル高速化百選
モデル高速化百選
 
“Understanding DNN-Based Object Detectors,” a Presentation from Au-Zone Techn...
“Understanding DNN-Based Object Detectors,” a Presentation from Au-Zone Techn...“Understanding DNN-Based Object Detectors,” a Presentation from Au-Zone Techn...
“Understanding DNN-Based Object Detectors,” a Presentation from Au-Zone Techn...
 
convolutional_neural_networks in deep learning
convolutional_neural_networks in deep learningconvolutional_neural_networks in deep learning
convolutional_neural_networks in deep learning
 
Neural network
Neural networkNeural network
Neural network
 
IOEfficientParalleMatrixMultiplication_present
IOEfficientParalleMatrixMultiplication_presentIOEfficientParalleMatrixMultiplication_present
IOEfficientParalleMatrixMultiplication_present
 
Tomographic reconstruction in nuclear medicine
Tomographic reconstruction in nuclear medicineTomographic reconstruction in nuclear medicine
Tomographic reconstruction in nuclear medicine
 

Último

Structuring Teams and Portfolios for Success
Structuring Teams and Portfolios for SuccessStructuring Teams and Portfolios for Success
Structuring Teams and Portfolios for Success
UXDXConf
 

Último (20)

Measures in SQL (a talk at SF Distributed Systems meetup, 2024-05-22)
Measures in SQL (a talk at SF Distributed Systems meetup, 2024-05-22)Measures in SQL (a talk at SF Distributed Systems meetup, 2024-05-22)
Measures in SQL (a talk at SF Distributed Systems meetup, 2024-05-22)
 
Speed Wins: From Kafka to APIs in Minutes
Speed Wins: From Kafka to APIs in MinutesSpeed Wins: From Kafka to APIs in Minutes
Speed Wins: From Kafka to APIs in Minutes
 
Behind the Scenes From the Manager's Chair: Decoding the Secrets of Successfu...
Behind the Scenes From the Manager's Chair: Decoding the Secrets of Successfu...Behind the Scenes From the Manager's Chair: Decoding the Secrets of Successfu...
Behind the Scenes From the Manager's Chair: Decoding the Secrets of Successfu...
 
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
 
Structuring Teams and Portfolios for Success
Structuring Teams and Portfolios for SuccessStructuring Teams and Portfolios for Success
Structuring Teams and Portfolios for Success
 
Syngulon - Selection technology May 2024.pdf
Syngulon - Selection technology May 2024.pdfSyngulon - Selection technology May 2024.pdf
Syngulon - Selection technology May 2024.pdf
 
ASRock Industrial FDO Solutions in Action for Industrial Edge AI _ Kenny at A...
ASRock Industrial FDO Solutions in Action for Industrial Edge AI _ Kenny at A...ASRock Industrial FDO Solutions in Action for Industrial Edge AI _ Kenny at A...
ASRock Industrial FDO Solutions in Action for Industrial Edge AI _ Kenny at A...
 
Custom Approval Process: A New Perspective, Pavel Hrbacek & Anindya Halder
Custom Approval Process: A New Perspective, Pavel Hrbacek & Anindya HalderCustom Approval Process: A New Perspective, Pavel Hrbacek & Anindya Halder
Custom Approval Process: A New Perspective, Pavel Hrbacek & Anindya Halder
 
SOQL 201 for Admins & Developers: Slice & Dice Your Org’s Data With Aggregate...
SOQL 201 for Admins & Developers: Slice & Dice Your Org’s Data With Aggregate...SOQL 201 for Admins & Developers: Slice & Dice Your Org’s Data With Aggregate...
SOQL 201 for Admins & Developers: Slice & Dice Your Org’s Data With Aggregate...
 
Choosing the Right FDO Deployment Model for Your Application _ Geoffrey at In...
Choosing the Right FDO Deployment Model for Your Application _ Geoffrey at In...Choosing the Right FDO Deployment Model for Your Application _ Geoffrey at In...
Choosing the Right FDO Deployment Model for Your Application _ Geoffrey at In...
 
ECS 2024 Teams Premium - Pretty Secure
ECS 2024   Teams Premium - Pretty SecureECS 2024   Teams Premium - Pretty Secure
ECS 2024 Teams Premium - Pretty Secure
 
Free and Effective: Making Flows Publicly Accessible, Yumi Ibrahimzade
Free and Effective: Making Flows Publicly Accessible, Yumi IbrahimzadeFree and Effective: Making Flows Publicly Accessible, Yumi Ibrahimzade
Free and Effective: Making Flows Publicly Accessible, Yumi Ibrahimzade
 
Introduction to FDO and How It works Applications _ Richard at FIDO Alliance.pdf
Introduction to FDO and How It works Applications _ Richard at FIDO Alliance.pdfIntroduction to FDO and How It works Applications _ Richard at FIDO Alliance.pdf
Introduction to FDO and How It works Applications _ Richard at FIDO Alliance.pdf
 
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
 
Connecting the Dots in Product Design at KAYAK
Connecting the Dots in Product Design at KAYAKConnecting the Dots in Product Design at KAYAK
Connecting the Dots in Product Design at KAYAK
 
Intro in Product Management - Коротко про професію продакт менеджера
Intro in Product Management - Коротко про професію продакт менеджераIntro in Product Management - Коротко про професію продакт менеджера
Intro in Product Management - Коротко про професію продакт менеджера
 
PLAI - Acceleration Program for Generative A.I. Startups
PLAI - Acceleration Program for Generative A.I. StartupsPLAI - Acceleration Program for Generative A.I. Startups
PLAI - Acceleration Program for Generative A.I. Startups
 
Integrating Telephony Systems with Salesforce: Insights and Considerations, B...
Integrating Telephony Systems with Salesforce: Insights and Considerations, B...Integrating Telephony Systems with Salesforce: Insights and Considerations, B...
Integrating Telephony Systems with Salesforce: Insights and Considerations, B...
 
Top 10 Symfony Development Companies 2024
Top 10 Symfony Development Companies 2024Top 10 Symfony Development Companies 2024
Top 10 Symfony Development Companies 2024
 
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
 

Distributed target tracking under realistic network conditions

  • 1. Distributed target tracking under realistic network conditions Christian Nastasi Andrea Cavallaro c.nastasi@sssup.it andrea.cavallaro@eecs.qmul.ac.uk www.eecs.qmul.ac.uk/~andrea/wise-mnet.html
  • 2. Context • Wireless Sensor Networks (WSNs) – networks of autonomous sensors used for pervasive applications – large-number deployments, highly scalable – resource-constrained – scalar data (e.g. temperature, light, pressure) The way ahead... • Wireless Multimedia Sensor Networks (WMSNs) – vectorial data (e.g. audio, video) – raw data cannot (always) be transferred – local processing is required (but much more complex!) C. Nastasi, A. Cavallaro Distributed target tracking under realistic network conditions
  • 3. Application: multi-sensor tracking • Objective Continuous estimation of the target state given a set of measurements (observations) obtained from spatially distributed sensing nodes. 4 zk Z k  ( zk 1 2 N zk  zk ) 1 zk Measurements 2 zk xk  f(Z k , Z1:k 1, x0:k-1 ) 3 State estimation zk C. Nastasi, A. Cavallaro Distributed target tracking under realistic network conditions
  • 4. Approaches Centralized Decentralized Distributed Distributed and decentralized multi-camera tracking M. Taj, A. Cavallaro IEEE Signal Processing Magazine, Vol. 28, Issue 3, May 2011 C. Nastasi, A. Cavallaro Distributed target tracking under realistic network conditions
  • 5. Distributed tracking: strategies • Distributed target tracking – need a collaborative information exchange mechanism – consensus-based algorithms • Parallel (e.g. Kalman Consensus Filter [Olfati-Saber2005], Distributed Particle Filters [Gu2007]) – data aggregation algorithms • Sequential (e.g. Distributed Particle Filters [Hlinka2009]) estimate start consensus aggregation C. Nastasi, A. Cavallaro Distributed target tracking under realistic network conditions
  • 6. Distributed Particle Filters (DPFs) • Basic ideas: – each node executes a local Particle Filter (PF) – measurements are synchronized, calibration is known – some information is exchanged • Likelihood sharing [Coates2004] – exchange information to have a common model of the likelihood – random number generators are synchronized • Posterior sharing – the network has a common knowledge of the posterior pdf – consensus-based approach [Sheng2005, Gu2007] – aggregation-based approach [Sheng2005, Hlinka2009] • spatial sequence of aggregation steps • Partial Posterior (PP) is exchanged among the nodes C. Nastasi, A. Cavallaro Distributed target tracking under realistic network conditions
  • 7. Aggregation-based DPF f(x k | Z1:k 1 , z k:4 )  f(x k | Z k ) 1 1 f(x k | Z1:k 1 , z k ) start estimate 4 zk 1 zk 2 zk f(x k | Z1:k 1 , z k:2 ) 1 3 zk f(x k | Z1:k 1 , z k:3 ) 1 Problem: Particle dissemination is not feasible! Solution: Gaussian Mixture Model of the Partial Posterior (GMM-PP) Independence from the # of particles C. Nastasi, A. Cavallaro Distributed target tracking under realistic network conditions
  • 8. How to extend this tracking approach from WSNs to WMSNs?
  • 9. Proposed approach • Objective – Distributed tracking under realistic conditions in camera-based WMSNs • Problems – existing approaches are theoretical and designed for WSNs – need adaptation for limited Field-Of-View sensors (cameras) • detection miss • target hand-over • target loss – need mechanisms for the definition of the aggregation chain • first node (starts iteration) • intermediate nodes (aggregate local measurement to the PP) • last node (performs estimation) – a network-simulator environment is required C. Nastasi, A. Cavallaro Distributed target tracking under realistic network conditions
  • 10. First node x (i) k -1 ,w  (i) P k -1 i 1 1. Knows previous posterior 1 z k and local measurement 2. Prediction and Update: • re-sampling • draw from state-transition • weight update from likelihood next-hop 3. GMM-PP creation x  (i) (i) P 4. Next-hop selection k -1,w k -1 i 1 5. Sends GMM-PP x (i) k ~ f(x k |x (i) k -1 ) i  1,..., P 1 f(z k|x k ) (i) x (i) k ,w  (i) P k i 1 f 1 GMM - PP w (i) k  P 1 (j) i  1,..., P  f(z k|x k ) C. Nastasi, A. Cavallaroj 1 Distributed target tracking under realistic network conditions
  • 11. Intermediate node h next-hop 1. Receives PP from node h-1 2. Importance sampling: • use the incoming PP as importance function g() • draw from importance function • weight update: CONDENSATION h zk 3. GMM-PP creation 4. Next-hop selection g(x k )  f 1:h -1 GMM - PP (x k | z 1:h -1 k ) 5. Sends GMM-PP x (i) k ~f GMM - PP (x k | z k:h -1 ) 1 i  1,..., P w (i) k  P h f(z k|x k ) (i) i  1,..., P x (i) k -1 ,w  (i) P k -1 i 1 1:h f GMM - PP h (j)  f(z k|x k ) j 1 C. Nastasi, A. Cavallaro Distributed target tracking under realistic network conditions
  • 12. Last node First node at k+1 1. Receives PP from node N-1 2. Importance sampling as for N zk intermediate nodes 3. Last PP is also the global PP 4. Target state estimation 5. Next tracking step starts here! After importance sampling: f PPN  f(x k | Z k ) 1: P Estimation: x k   wk x k ˆ (i) (i) i 1 C. Nastasi, A. Cavallaro Distributed target tracking under realistic network conditions
  • 13. Experimental setups • Simulations • number of nodes: N = 10, 50, 100, 300, 500, 700, 1000 • number of particles: P = 100, 300, 500 • DPF with different GMM configurations • No GMM approximation: DPF-0 • Variable number of GMM components: DPF-1, DPF-5 • realistic network conditions Simulator: WiSE-MNet www.eecs.qmul.ac.uk/~andrea/wise-mnet.html C. Nastasi, A. Cavallaro Distributed target tracking under realistic network conditions
  • 14. Simulation setup • Network – T-MAC protocol, BW = 250 kbps – request-to-send/clear-to-send mechanism – acknowledged-transmission mechanism – number of retransmissions: 10 • Cameras – Covering 6000 sqm (random uniform distribution) – Top-down facing cameras: 6m from the ground plane (FOV is 10m X 6m) – Frame rate = 1fps • 100 simulation runs, each of 10 minutes C. Nastasi, A. Cavallaro Distributed target tracking under realistic network conditions
  • 15. What do we measure? K tr Ktr # of estimations (detected events) • Estimation efficiency E K K # of observations (all the events) K tr • Average estimation delay D  1  d(k) K tr i 1 d(k) : Estimation delay for the k-th tracking step C. Nastasi, A. Cavallaro Distributed target tracking under realistic network conditions
  • 18. Conclusions • Conclusions – distributed target tracking for camera-based WMSNs with a DPF • Dealing with limited-FOV sensors • Operating on a network-simulator environment – importance of co-design between tracking algorithms and communication protocols Simulator available as open source at www.eecs.qmul.ac.uk/~andrea/wise-mnet.html • Future work – Comparing other state-of-the-art protocols (e.g. consensus-based) – Using the full vision-pipeline: more complex features C. Nastasi, A. Cavallaro Distributed target tracking under realistic network conditions