1. Distributed Clustering for Scalable, Long-Lived Sensor Networks
Ossama Younis and Sonia Fahmy
Department of Computer Sciences, Purdue University
1. Model
2. System Challenges
3. Clustering Challenges
How can the system provide:
Scalability (to thousands of nodes)?
Prolonged network lifetime?
Data and state aggregation?
Robustness in the face of unexpected failures?
Security of sensor communications?
Approach Clustering
Completely distributed
Energy-efficient
O(1) iterations to terminate (i.e., independent of network diameter)
Low message/processing overhead
High energy, well-spread cluster heads
Other cluster characteristics, such as balanced or dense clusters
Approach HEED (Hybrid, Energy-Efficient, Distributed clustering)
Network:
Rectangular field with a large number of dispersed sensor nodes
Sensor nodes:
Location un-aware and quasi-stationary
Homogeneous and equally significant
Unattended (infeasible to recharge)
Example applications:
Seismic monitoring or field surveillance.
4. Parameters
5. HEED at node v
6. Examples
Primary parameter (maximize energy)
Residual Energy (Er)
Secondary parameter (minimize cost)
Used to break ties
Balanced clusters
Discover neighbors(v) within cluster range
, 0 <Cprob< 1, Emax = max. battery power, CHprob ≥ pmin (e.g., pmin = 10-4)
CH: Cluster Head
Cost definition N Y
CHprob≤1 ?
node degree
(for load balancing)
AMRP: Average min. reachability power (for min. intra-cluster comm. energy)
1/node degree
(for dense clusters)
Dense clusters
Covered?
Covered? N N Y
v is CH Y
7. Properties
received
CH msg?
Completely distributed √
Clustering requires,
i.e., O(1) iterations √
Message overhead: O(1) per node √
Processing overhead: O(N) per node √
Pr{two CH’s within the same cluster range},
,where p=initial CHprob Y N
Stop
Clustering 1000 nodes shows the least variance in cluster size for the load balanced organization
Pick CH with lowest cost
Elect to become CH with prob. CHprob
CHprob = CHprob x 2
8. Sample Performance Results
9. Now what ?
How effective is HEED when integrated with a hierarchical routing organization?
Is it possible to extend HEED to serve networks operating in environments where unexpected failures occur?
Yes, we are currently designing the READ (Robust, Energy-Aware, Distributed) clustering protocol. READ can generate network graphs that are k-fault tolerant by building multiple independent cluster structures.
How can HEED be adapted to secure sensor communications?
Contact us:
For more information:
Compare HEED clustering quality to a weight-based clustering approach (e.g., DCA). Setup: 2000x2000 area, 1000 nodes, node residual energy = Uniform(0,1) Joule
Apply to an application and compare to an optimized version of LEACH (Heinzelman et al., 2002) and direct communication to demonstrate how clustering prolongs the network lifetime
HEED selects CHs that are rich in energy
Energy dissipates slowly as the distance increases
HEED overhead is not significant