Localized Low-Power Topology Control Algorithms in IEEE based Sensor Networks Jian Ma *, Min Gao *, Qian Zhang +, L. M. Ni *, and Wenwu Zhu + * Department of Computer Science, Hong Kong University of Science and Technology + Wireless and Networking Group, Microsoft Research Asia IEEE International Conference on Distributed Computing Systems (ICDCS), 2005 Wang, Sheng-Shih May 26, 2005

Outline Introduction Proposed Pruning Algorithms Simulation Conclusion

Introduction --- IEEE Node roles PAN coordinator Coordinator Device Device types FFD (Full Function Device) RFD (Reduced Function Device)

Introduction --- IEEE Star topology Peer-to-peer topology

Introduction --- IEEE Indirection data transmission coordinatordevice Beacon Data Request Acknowledgement (optional) pending list Data Acknowledgement The transaction is instigated from the device The device need turn on its receiver when its coordinator sends beacons  Save power ModePower consumption Transmit36.5 mW Receive41.4 mW Idle41.4 mW sleep 42  W

Introduction --- IEEE based Sensor Networks Sink PAN coordinator Features Powerful computing capability Unlimited power supply Sensor Coordinator or device

Introduction Challenge How many coordinators are required ? Objective Lower power consumption by choosing as few coordinator as possible

Pruning Algorithm ? OK C doesn ’ t know when B will turn on B is a device B ’ s coordinator is A Theorem: A route between two nodes is bidirectionally connected if and only if all intermediate nodes along the route are coordinators

Pruning Algorithm --- Problem Formulation Finding a minimal connected coordinator set (i.e., backbone) in a WSN Pruning techniques Self pruning algorithm Ordinal pruning algorithm Layered pruning algorithm

Pruning Algorithm --- Assumptions All nodes are stationary FFDs No isolated node For each node The node is either a coordinator or associated with one of the coordinators Has the shortest hop distance to the sink Has 2-hop neighbor information

Self Pruning --- Concept Main idea The nodes near the sink always need forward the traffic between the sink and nodes far from the sink Definition: priority Node u has a higher priority than node v if d ( u ) < d ( v ) or d ( u ) = d ( v ) AND (node u has a smaller ID)

Self Pruning --- Rule Let S(v)=N(v)  {u|u has a higher priority than v} Connectivity test Coverage test

Self Pruning --- Example sink N (3) = {0, 1, 2, 7, 8} Nodes 7 and 8 both have lower priorities than nodes 0, 1, and 2 S (3) = {0, 1, 2}  nonempty and connected Node 7 is node 1’s neighbor Node 8 is node 2’s neighbor  Node 3 acts as a device

Self Pruning --- Summary Every node makes its decision independently Maintain the network connectivity Keep the shortest path between each node and the sink Running time: O(1)

Ordinal Pruning --- Concept Reduce the number of coordinators Self pruning does not consider the possibilities that some neighbors with lower priorities become coordinators More neighboring coordinators could increase the chance to pass the coverage test

Ordinal Pruning --- Rule Let S(v)=N(v)  {u|u has a higher priority than v or u has a lower priority but becomes a coordinator}. Node v will not be a coordinator if Each node decides its role only after all its neighbors with lower priorities have determined their roles

Ordinal Pruning vs. Self Pruning S(5)={1, 2, 4}  coordinator S(5)={1, 2, 4, 6, 7}  device ? V

Ordinal Pruning --- Summary The shortest path is not guaranteed All nodes sequentially decide their roles Running time: O(n)

Layered Pruning --- Concept Tradeoff between the number of coordinators and running time Each node just considers all neighbors with higher hop distances All nodes in the same layer (i.e., same hop distance) decide their roles simultaneously

Layered Pruning --- Rule Let S(v)=N(v)  {u|u has a higher priority than v or d(u) > d(v) but u becomes a coordinator} Node v will not be a coordinator if

Layered Pruning --- Example Nodes 9, 10, 11, 12, 13, 14, and 15 determine their roles in the first step

Layered Pruning --- Summary Running time depends on the hop distance of the farthest nodes Grid topology:

SP vs. OP vs. LP Self Pruning Ordinal PruningLayered Pruning 7 coordinators5 coordinators6 coordinators S(2)={0, 1}  Coverage test is failed (node 6) S(2)={0, 1, 4, 6, 7}  Connectivity test is failed S(2)={0, 1, 4, 5, 6, 7}  Connectivity test passed  coverage test passed

Simulation --- Node Placement Uniform Grid topology All nodes are evenly divided amongst the grids 49 nodes (sink is placed at the center) Random Network size: 100m  100m Communication range: 10m Node number: 200 ~ 2000

Simulation --- Uniform Placement

Simulation --- Random Placement

Conclusion Less average end-to-end delay  Self pruning algorithm Power saving  Ordinal pruning algorithm Different algorithms are used in the different networks based on their application requirements

Future Works Role rotation policy Balance energy consumption among all nodes Consider the remaining power Dynamic reconstruction of coordinators Handle the failure of some coordinators