Query Evaluation in Wireless Sensor Networks Project Presentation for Comp 8790 Student: Yongxuan Fu Supervised by: Prof. Weifa Liang Presented on: 07/11/13
Outline Background Preliminary Algorithm Design Simulation and Experimental Evaluation Conclusion Question
Background Wireless Sensor Networks Top K query Dynamic Transmission Route
Background - WSNs
Background Top-K query it is one of those fundamental ones that focus on looking for the most important events sensed by the network and judged by ranking methods. It requires the response from all sensor nodes It is also quite useful in many applications
Background Dynamic Transmission Route Release those nodes with lower battery level from heavy transmission relay duty and assign extra duty to the nodes with higher energy level. With a cost of extra energy consumption One way that balance the between minimizing of total energy consumption of the network and reducing worst individual energy consumption
Preliminary Problem and Aims Naive-K Optimal Quantile Filter(QF) Energy-Efficient Dynamic Routing Tree Algorithm (EDRT)
Preliminary - Problem and Aim Problem: Severe constrains of limited and low capacity of energy of sensor nodes Aim: Maximize the lifetime of WSNs
Preliminary Naive-K A widely used simple Top-K query evaluation Algorithm Step1: Every sensor node send K largest value data points to its parent node Step2: Base Station receives n * K data points from its n direct children nodes and calculate the query result
Preliminary - QF This is a novel filter-based algorithm for energy-efficiency topk query evaluation. Step 1: a node sort all sensed and received data points in decrease order and send one data point to parent node according to pre-set alpha Step 2: parent node calculate the value of filter and send back to all its children node. Step 3: child node send all data points that have the value no less than the filter as the result of the query within its branch to the parent node.
Preliminary EDRT This is a novel algorithm proposed for dynamic adjustment of WSN routing tree. It is a query dependent algorithm It randomly choses the route from candidates It classified all nodes into deferent level according to transmission distance to minimize the depth of routing tree.(this inspired our proposed algorithm)
Algorithm Design This is a novel algorithm proposed by us for dynamic adjustment of WSN routing tree that can be used in more applications than EDRT. It is query independent It choses route according to candidates energy level It is a locally adjustment algorithm It implements the level of distance concept to minimize the depth of routing tree and to avoid local close loop
Algorithm Design cont. Three Stages of the algorithm Stage 1: Pre-set the energy checking thresholds Stage 2: Initial the distance level of every nodes in the network and connect Stage 3: maintain the network
Simulation and Experimental Evaluation Simulator Experiment 1: Impact of k on Performance of the networks Experiment 2: Impact of Density of nodes on Performance of the networks
Simulator Developed in Java (jre 1.7) using eclipse Kepler Can simulate a n * m region of interest with one base station and j sensor nodes with the uniform transmission range L The network can run in 4 different mode Naive-K with static minimum depth routing tree QF with static minimum depth routing tree Naive-K with dynamic routing tree QF with dynamic routing tree It will report the network s lifetime
Experiments 1 result 7000 6000 5000 4000 3000 2000 1000 0 100 150 200 250 300 350 400 450 500 Impact of the value of k on different networks (Single run) QF with dynamic route QF with static route Average lifetime with multiple runs of each scenario
Experiment 2 result Network Type number of sensor nodes network lifetime 1000 1117 Naive-k with dynamic route 2500 1071 5000 931 1000 1143 Naive-k with static route 2500 931 5000 855 1000 6490 QF with dynamic route 2500 5889 5000 5706 1000 4993 QF with static route 2500 4280 5000 4103
Results 1: The proposed algorithm can really help maximizing networks lifetime 2: The proposed algorithm can co-operate with complex query evaluations 3: The value of K has no significant impact on the performance of the proposed algorithm 4: The performance of the proposed algorithm will be low while the density of sensor is high. 5: the QF algorithm outperforms Naive-K significantly in all scenarios
Conclusion Proposed a novel Dynamic Routing Tree algorithm Evaluate the performance of the algorithm and the impacts of parameters on it. Examined two different Top-K query evaluation techniques while co-operate with our proposed algorithm
Question?