ECE 555 Mid-term Presentation Two-radio Multi-channel Real-time

Document Sample
ECE 555 Mid-term Presentation Two-radio Multi-channel Real-time Powered By Docstoc
					ECE 555 Mid-term Presentation

Two-radio Multi-channel Real-time Wireless Sensor
Networks

by Xiaodong Wang, Yanjun Yao

Oct 20th, 2008
         Project Outline

      Purpose:
       Develop a real-time wireless sensor networks for the
       application of data-centre temperature monitoring
            High Density Network
            High workload
            Power is not a big concern

      Approach
            Using multi-radio multi-channel wireless sensor networks to increase the
             parallelism




      Mid-term Plan
            Find out the constraints of minimizing the sums of end-to-end delay of all
             nodes
            Develop the channel allocation algorithm for the system



Page 2
         System Model

          Directed network graph G=(V,E)
           with weighted edge




          Node channel allocation vector
                                 
                       ci u   c1ui , c 2ui ,...,ct ui   
                                                           '
                                                                            
                                                                 co u   c1uo , c 2 uo ,...,c t uo   
                                                                                                       '



                                        c k ui  1         If channel k is assigned

          Edge weight vector
                          w w 2,...,
                            1, uv w
                            euv  w t
                                uv
                                     '
                                    uv                            
                          w  /PRRif c k uo  c k vi  1
                           k
                            uv 1 uv




Page 3
         System Model (cont’)
          Node weight vector[1]

                                   vw v
                                 v 1 v t  
                                 ,2 w 
                                 ww,..., w euv           '

                                                              
                                                              V,
                                                              ueE
                                                               uv



          Example:
                                                      c  red, black , blue 
                                                                             '




                                                   ci B  1,1,0 co B  1,0,1
                                                                   '                '




                                                   wB  we FB   we EB   we AB 
                                                          3,0,0'4,4,0'0,10,0'
                                                          7,14,0'




          [1] “Flow-Based Real-Time Communication in Multi-Channel Wireless Sensor Network”,
               Tech Report, EECS Dept, University of Tennessee, Knoxville, 2008


Page 4
         Problem Formulation and Optimization Goal
          Define node delay:
                                        du   max{w i u }

          Goal: try to find feasible ci u  and co u  for every node such that
                                        min max du 
                                                  uV

          Constraints:
                                u, v  u , co u   ci v   2
                                            t                 t
                                 u V ,  c      k
                                                      uo     c k ui  2
                                           k 1              k 1




Page 5
         Channel Allocation Heuristic Design

          Establish a fat tree by using Breadth-First search algorithm




Page 6
         Channel Allocation Heuristic Design (cont’)

          Channel assignment of Multi-radio
             We allocate the channels for node by level sequence.
             Every node chooses the minimum used channel as its own downstream
              channel.
             the lower level node decide which parent node to join.


          Metric used for choosing parents
                                 
                              __  
                             Current i
                              Path
                               Delay
                                  i Influen            
          Example




Page 7
         Implementation

      We simulate the algorithms on our own implementation code written
       in c++.

      Baseline:
           Single channel, single radio.
           Same method of building fat tree, different rule of choosing father node.


      Compare on the sum of path delay of all nodes, on cases:
           The sparse topology:
              121 node case -----   few communication links, few interference links
              169 node case -----   few communication links, no interference links
              225 node case -----   medium communication links, few interference links
           The dense topology:
              289 node case -----   medium communication links, large interference links
              361 node case -----   large communication links, large interference links



Page 8
         Result for Sparse Topology
          Y-axis: u VPath u
                       _delay
                    
          SR-1C: Single Radio Single Channel (baseline)
          MR-nC: Multi Radio n Channel
               9000
               8000
               7000
                                                           SR-1C
               6000
                                                           MR-2C
               5000
                                                           MR-3C
               4000                                        MR-4C
               3000                                        MR-5C
               2000                                        MR-6C
               1000
                  0
                      121 nodes   169 nodes 225 nodes


Page 9
      Result for Dense Topology
       Y-axis: u VPath u
                    _delay
                 
       SR-1C: Single Radio Single Channel (baseline)
       MR-nC: Multi Radio n Channel
               30000

               25000
                                                        SR-1C
               20000
                                                        MR-2C
               15000                                    MR-3C
                                                        MR-4C
               10000                                    MR-5C
                                                        MR-6C
                5000

                     0
                         289 nodes       361 nodes



Page 10
      Plan

           Final goal
              Finding a optimal or sub-optimal solution (algorithm) for the minimization
               problem
              Develop the simulation platform in NS-2
              Do simulation and compare with other real-time protocol


           If time allows
              Divide nodes into clusters, and estimating the missing data according to the
               data we get.
              Dynamically change the resource allocation based on the network condition
               and data importance to achieve local optimal




Page 11
      Work Division

           Problem formulation: Xiaodong
           Implementation: Yanjun, Xiaodong
           Heuristics Design: Xiaodong, Yanjun




Page 12
      Q&A




Page 13