Coverage Protocol for Wireless Sensor Networks Using Distance Estimates Mingze Zhang, Mun Choon Chan and A. L. Ananda School of Computing National University of Singapore, IEEE SECON 2007 26 OCT 07 Gilsoo Kim

Contents Introduction Neighbor node distance estimation Problem formulation Maximum likelihood distance estimation Evaluation Configurable Coverage Protocol (CCP) Vacancy inside triangle Node selection constraints Protocol description Conclusion

Introduction Coverage problem in WSNs Motivations Failure-prone sensor devices are deploy in higher density to meet various design specifications Number of active sensors needed to cover the area of interests is to be minimized Motivations Global localization is expensive, error-prone and not required for coverage algorithm By requiring complete coverage, excessive overlap may occur

Contents Introduction Neighbor node distance estimation Problem formulation Maximum likelihood distance estimation Evaluation Configurable Coverage Protocol (CCP) Vacancy inside triangle Node selection constraints Protocol description Conclusion

Problem formulation Estimating the distance b/w two nodes can be easier and less error-prone than global localization (na-nx), (nb-nx), and nx are correlated By taking into account values of them, d can be estimated na nb nx

Maximum likelihood distance estimation Maximum likelihood estimation is used to estimate the size of X and thus the distance d Node distribution can be estimated as a Poisson (random) point process Probability of having certain number of nodes inside an area given the value of the area where,

Maximum likelihood distance estimation Additional information helps to improve the estimation accuracy Multiple transmission power levels Example Two power level sensor nodes Final estimates can be calculated as the average of four possible combinations

Evaluation (1/2) Compare the performance of using one and two transmission power levels Power levels are 0.5 and 1 (normalized values) Multiple Tx power gives significant improvements Performance improves with the increasing node density Distance estimates can provide enough accuracy for coverage problem

Evaluation (2/2) Adaption of realistic propagation model Upper and lower bound Degree of irregularity (DOI) Maximum radio range variation per unit degree change in the direction of radio propagation DOI=0.05 DOI=0.2 Estimation error increases almost linearly with DOI Average error can still be tolerable for coverage applications (about 15% of rc) Two power level: First=0.25~0.5, Second=0.5~1

Contents Introduction Neighbor node distance estimation Problem formulation Maximum likelihood distance estimation Evaluation Configurable Coverage Protocol (CCP) Vacancy inside triangle Node selection constraints Protocol description Conclusion

Configurable Coverage Protocol (CCP) CCP only makes use of the distance information among the neighboring nodes Can be built on any other distance estimation scheme CCP allows the users to specify the coverage objective α The ratio of the size of the vacancy inside the triangle to the area of triangle should be less than or equal to 1- α By ensuring that coverage objective is met locally, the global coverage will be satisfied too

Vacancy inside triangle (1/2) Triangle vacancy calculation Area of triangle (Heron’s formula) Overlapping area b/w two nodes

Vacancy inside triangle (2/2) Exceptional cases of vacancy calculation CCP tries to avoid exceptional cases during selection of active nodes Hard to calculate Potentially increases the number of active nodes For a sufficient high node density, these cases are not likely to happen Even when these cases are included and no vacancy is assumed the error is small Exceptional cases of triangle vacancy calculation Inefficiency caused by exceptional cases

Node selection constraints (1/2) Connectivity constraints Set of active sensor nodes must be connected at all times A node should only volunteer itself if it is able to communicated with both end vertices of the edge Angle constraints Exceptional cases analyzed in previous section shall be avoided as much as possible Exceptional cases occur when there are small (or large) angles inside the triangle CCP tries to select the triangle that maximizes the minimum angle

Node selection constraints (2/2) Angle constraints (cont’d) Any node that has an angle smaller than ß1 will just ignore the new triangle and edge message It is essential for every node that is trying to compete for new vertex to hear power on message Node that can form an angle larger than max(ß1, ß2) meet the angle constraints and are preferred

Protocol description (1/3) Selection of starting node At the beginning, all nodes are in “UNDECIDED” state A node should volunteer to be starting node with probability p (normally p=1/N) It first waits for a random time ts=[0, tsmax] If it doesn’t hear any messages within ts, it will change its state “ON” and broadcast the power on message If it receives any power on message, it will simply cancel the timer

Protocol description (2/3) First edge and first triangle formation First edge All neighbors around starting node will set a timer t1 If the timer is fires, the node will change its state to “ON” When a node turns “ON”, it broadcasts power on message including edge information (IDs, length of edge) First triangle Upon receiving the edge information, the neighboring node will set a timer t2 If the timer fires, the node turns “ON” and form the first triangle The node will broadcast the power on message including the triangle information (IDs, length of three edges)

Protocol description (3/3) Node selection process Each node will first examine whether it has any triangle associated w/ itself and share a common vertex If there is only one triangle associated with the edge, and it satisfies the vacancy requirements, it will announce an creation of a new triangle If there are already two triangles associated with this edge, it will take no action When a node notices that it is within one of the triangles formed, it turns itself “OFF” The protocol terminates when all nodes are either in the “ON” or “OFF” states

Evaluation (1/3) Performance metrics Parameter settings Average vacancy Number of active nodes Parameter settings Parameters value Sensing radius rs 1 (normalized value) Communication range rc 3 Area size 30 x 30 Two power level rc=1, rc=2 a, b in time t2 0.5

Evaluation (2/3) Comparison between CCP and ODGC (α =1) ODGC has a slightly better performance However, performance degradation is small The vacancy is a result of the distance estimation and location error Only coverage objective of 0.98 or below can be achieved

Evaluation (3/3) Performance of CCP with α <1 CCP is able to meet the coverage objectives most of the time Reasons why the objective can not be met Network density is too low Distance estimation error Biggest savings comes from moving from 95% to 90% coverage (12%)

Conclusion Simple distance estimation scheme Based on number of node Configurable coverage protocol Needs only distances among the neighboring nodes Vacancies are estimated distributively Global coverage objective α can be maintained CCP performs very similar to OGDC for complete coverage Less active sensor nodes are needed than necessary amount

Q & A Thank you