Energy Efficient Spanning Tree for Data Aggregation In Wireless SENSOR NETWORKS

1. Energy Efficient Spanning Tree for Data Aggregation In Wireless SENSOR NETWORKS Mohammad Tariqul Islam (Tarik)

2. The BEAST (Problem) & the BEAUTY(SOLUTION) • WSN sensor nodes have limited, non-renewable energy source. • Need to preserve energy to maximize lifetime • Aggregate data at the intermediate nodes to: • Eliminate redundancy • Reduce size of the data to be transmitted ( less spent energy ) • Aggregation Spanning Tree approach: • Construct a spanning tree from the connectivity graph • Send data along the path to the root • Data is aggregated at the intermediate nodes towards the root.

3. ROADMAP • GENERAL DISCUSSION • RELATED WORK • PROTOCOL DESCRIPTION • EVALUATION

4. HOW DOES DATA AGGREGATION HELP US? • Energy consumption for transmission has direct correlation with distance between sender and receiver. • Collision resolution is delayed in case of simply stacked on packets or separate transmission without data aggregation • BUT FOR AGGREGATING PACKETS I HAVE TO WAIT FOR LONGER TIME!!! • Delay for waiting is negligible for sensor nodes with shared sensing area. • WHAT ABOUT ENERGY COST, HUH? DON’T I HAVE TO USE UP MY ENERGY FOR AGGREGATION? • Chill – consumed energy for data aggregation is much less than transmitting own packet.

5. HOW DOES IT WORK? • All nodes are in sleep mode when there is no data to be sent. • Event = > nodes sensing this event wakes up, senses data, and send it to the sink. • Data forwarding follows the spanning tree towards the root (sink) • Each intermediate node aggregates data and forwards it to its parent. • How do we evaluate the spanning tree? • Transmission delay - depends on the path depth • Scheduling policy • Queuing delay

6. Related works • ESPAN : ENERGY AWARE SPANNING TREE ALGORITHM • Uses nodes sensing the event region to perform data aggregation • Node with highest residual energy is chosen as the root • Other sensor nodes select their parent from the neighbors which has the shortest distance from the root. • Tie is broken from multiple candidates by choosing the node with highest residual energy. • Drawback: a node can be chosen as parent to many, leading to its energy exhaustion • LPT: • Node with highest residual energy is chosen as the root • Other sensor nodes select their parent from the neighbors which has the highest residual energy. • Does not consider distance to root in parent selection

7. EXAMPLE OF ESPAN PROTOCOL

8. EXAMPLE OF LPT PROTOCOL

9. EXAMPLE OF PROPOSED PROTOCOL

10. PROPOSED PROTOCOL • Selects the node with the highest residual energy as the root. • Each other node selects their parents by considering highest average path residual energy. • In case of same average path energy, node with the shortest distance to root is selected. • Moreover, a node can have at most a predefined (depending on node distribution and energy coverage) number of children. • Difference with ESPAN: Does not need to keep information about all the nodes, thus scalable.

11. PROPOSED PROTOCOL (CONTD.) • Average Path Energy: • sum of energy of the nodes along the path to the root (except the root) / path length • The Protocol: • Sr : sent by root, contains root ID • Sn : sent by each other sensors n, contains n. res_energy, n.distance_to_root, path energy, n.ID • Initialization: • Calculate max_child_count ( based on network density) • For all nodes i do: • i.ID = i • i.parents_distance_to_root = Inf • i.parent_energy = 0 • i.child_count = 0

12. PROPOSED PROTOCOL (CONTD.) • The Protocol (contd.): • Build Aggregation Tree: • Node with highest residual energy is selected as the root. • Root broadcasts Sr packet containing its ID. • Each of the roots neighbors n selects root as parent and sets • n.parent_ID = Sr.ID • n.distance_to_root = 1 • If ( n.children_count < max_child) • Compute path energy • Broadcast Sn • A node (except for root) receiving Sn • If(senders path energy avg > parents path energy avg) • Select sender as new parent

13. Evaluation

14. Evaluation (CONTD.)

15. Evaluation (CONTD.)

16. Energy efficient geographical forwarding algorithm for wireless sensor networks Mohammad Tariqul Islam (Tarik)

17. ROADMAP • NETWORK MODEL • PROTOCOL DESCRIPTION WITH EXAMPLE • ENERGY CRITICALITY CONSIDERATION

18. Network Model • Network is represented by a graph G=(V,E) • Each neighbor can estimate required power to transmit to a neighbor • Power value for a path: sum of power values of its constituent links • Each node knows its exact location information • Each node knows its one hop topology – neighbors and their energy (via periodic HELLO)

19. Protocol Overview • Overall idea: • Apply localized Djikstra’s Shortest Path algorithm to find best next hop • Repeat for every hop until the sink has been reached • Excludes nodes that are in critical energy state • Energy levels are broadcasted by every node after deployment and when drops below any critical level • Energy levels are expressed as an integer between 0 and L ( small positive integer) • Criticality denoted at multiple percentages, r1, r2, …, rk meaning ri fraction of nodes with lowest energy levels are in critical condition ri

20. PROTOCOL DESCRIPTION • For node that has new data, construct local topology G’(x) • Cost for a direct edge (u,v) where v is not the destination: • P(u,v) = (a*duva+c)/Eve • Cost of link (u,t):

21. EXAMPLE

22. Applying energy criticality • Define energy criticality percentage levels: • r1, r2, …, rk • When applying localized Djikstra’s algorithm • First use only those nodes from G’(x) which are above critical level indicated by r1 • If no next hop found, apply r2, and so on. • If no best hop found at all, flood the packet to overcome local maxima.

23. END