Multi-Criteria Routing in Pervasive Environment with Sensors Santhanakrishnan, G., Li, Q., Beaver, J., Chrysanthis, P.K., Amer, A. and Labrinidis, A Department of Computer Science University of Pittsburgh U.S.A. International Conference on Pervasive Services, (ICPS '05) Chien-Ku Lai
Outline Introduction Multi-Criteria Routing Protocol Performance Evaluation Conclusions and Future Work
Introduction 1. Wireless sensor networks (WSNs) 2. The major challenge in WSNs 3. The contributions of this paper
Introduction - Wireless sensor networks (WSNs) Sensor networks will be an integral part of a pervasive computing environment Since they allow interaction with the physical environment
Introduction - The major challenge in WSNs Power conservation Communication costs Network processing
Introduction - The major challenge in WSNs (cont.) In-network processing To perform computation in the network itself Reducing the size of the data to be sent higher up to other nodes Helps in reducing power consumption Since computation is cheaper in terms of energy and power than communication
Introduction - The major challenge in WSNs (cont.) More and more approaches adopting in- network processing of data The creation of the routing tree Base on the semantics of the query Energy remaining Power consumption model
Introduction - The contributions of this paper The introduction of a semantic and multi-criteria based routing protocol Self-optimizing Performance improvements Network lifetime Network coverage Survivability of critical nodes
Multi-Criteria Routing Protocol 1. Credit-Based Dynamic Route Update 2. Neighborhoods and Criteria Lists 3. Updating Credits 4. Proportional Credit Updates
Multi-Criteria Routing Protocol Tree structure Traditionally, signal strength is the main factor
Multi-Criteria Routing Protocol Current System State (Overall) Goal to be Satisfied by the System (eg. Network Coverage of 50% Multi-Criteria Algorithm (Per-node) Multi-Criteria Algorithm (Per-node) Criteria Pool (Energy Remaining, Power Consumption mode, etc.)
Multi-Criteria Routing Protocol
Credit-Based Dynamic Route Update The construction of the routing tree starts with a tree build request Initiated by the root node An identifier for the sender The query specification A value representing the current level in the tree level, L(sender)
Credit-Based Dynamic Route Update (cont.)
For selecting a node ’ s parent Power consumption model per node Watts Energy remaining at nodes Joules The group membership information For in-network aggregation Spatial locality Temporal locality
Neighborhoods and Criteria Lists
Updating Credits A set of goals are defined initially Initially the credits are distributed uniformly The base station updates credits among criteria Depending on the observed outcome
Proportional Credit Updates The redistribution of credits is done globally Checking periodically if the goal is satisfied The credits are redistributed proportionately The network is reconfigured
Performance Evaluation 1. Experimental Setup and Workload 2. Network Coverage 3. Network Lifetime 4. Survivability of Critical Nodes
Experimental Setup and Workload The simulator was written using C++ and csim The credit points were shaped from a pool of size 100 Various sensor network grid sizes from 15 x 15 to 50 x 50
Experimental Setup and Workload (cont.) Some standard SQL aggregation functions were used for the experiments SUM AVERAGE MAX
Network Coverage
Network Coverage (cont.)
Network Lifetime
Survivability of Critical Nodes
Conclusions and Future Work A multi-criteria routing scheme Minimal overhead Considering varied query frequencies, and varied (e.g., non-uniform) distributions of nodes
Questions? Thank you.