Design and Analysis of SD_DWCA - A Mobility based clustering of Homogeneous MANETs

This paper deals with the design and analysis of the distributed weighted clustering algorithm SD_DWCA proposed for homogeneous mobile ad hoc networks. It is a connectivity, mobility and energy based clustering algorithm which is suitable for scalable ad hoc networks. The algorithm uses a new graph parameter called strong degree defined based on the quality of neighbours of a node. The parameters are so chosen to ensure high connectivity, cluster stability and energy efficient communication among nodes of high dynamic nature. This paper also includes the experimental results of the algorithm implemented using the network simulator NS2. The experimental results show that the algorithm is suitable for high speed networks and generate stable clusters with less maintenance overhead.

💡 Research Summary

The paper presents a novel distributed weighted clustering algorithm, SD_DWCA (Strong Degree based Distributed Weighted Clustering Algorithm), designed for homogeneous mobile ad‑hoc networks (MANETs). The authors argue that existing clustering schemes—such as lowest‑ID, highest‑degree, mobility‑only, power‑only, or hybrid approaches—either ignore the quality of neighbor links, rely on unstable metrics, or overload cluster heads, leading to frequent re‑clustering and high maintenance overhead. To address these shortcomings, SD_DWCA introduces three key metrics for cluster‑head (CH) election: (1) Strong degree, defined as the number of “strong neighbors” that lie within half the transmission range (r/2) of a node; (2) Mobility metric, adopted from Xing et al., which computes a relative mobility value Rv(u) for each neighbor based on the change in Euclidean distance between two successive hello messages; and (3) Battery metric, which reflects the residual energy and recent consumption rate of a node. Each metric is normalized, weighted, and summed to produce a composite weight Wi for node i. Nodes with the highest Wi among their 1‑hop neighborhood become CHs; all other nodes join the CH with the highest weight they can hear.

The algorithm proceeds in two phases. In the initial clustering phase, every node periodically broadcasts hello packets containing its ID, position, and current weight. Upon receiving these packets, a node builds three neighbor sets: strong (SN), medium (MN), and weak (WN) based on distance thresholds (≤ r/2, r/2 < d ≤ 3r/4, 3r/4 < d ≤ r). The strong degree is simply |SN|, while the overall degree equals |SN| + |MN| + |WN|. Using the three metrics, each node computes Wi and decides whether to become a CH or to affiliate with a higher‑weight neighbor.

In the maintenance phase, at each hello interval nodes recompute their weights. If a node detects that its current CH’s weight has dropped below a predefined threshold or that a neighbor now has a higher weight, it may either re‑affiliate (switch to a different CH) or trigger a re‑clustering if the CH is no longer viable. The authors prioritize re‑affiliation over full re‑clustering to keep overhead low.

Complexity analysis shows that the initial clustering requires O(N·Δ) operations, where N is the number of nodes and Δ is the average degree, while maintenance incurs only O(N) per period, making the scheme scalable.

Simulation experiments were conducted using NS‑2 with node counts ranging from 50 to 200, transmission range 250 m, and maximum node speed up to 20 m/s. The authors compared SD_DWCA against four baseline algorithms: Lowest‑ID (LID), Highest‑Degree (HD), MOBIC (mobility‑only), and a previously proposed hybrid metric algorithm. Evaluation metrics included the number of clusters, average cluster size, CH change frequency, average end‑to‑end delay, and total energy consumption. Results indicate that SD_DWCA maintains significantly fewer CH changes (≈30 % of the baseline) even under high mobility, reduces overall energy consumption by about 15 %, and yields lower packet delay due to more stable backbone structures. The strong degree component effectively filters out weak and medium neighbors that are prone to move out of range, thus enhancing cluster stability.

The paper concludes that integrating strong degree, mobility, and battery metrics yields a clustering algorithm that balances connectivity, stability, and energy efficiency, outperforming existing schemes in dynamic MANET environments. However, the authors acknowledge that accurate distance measurement (e.g., GPS) is required to compute strong degree, and that GPS errors or heterogeneous transmission ranges could degrade performance. They also note that the weighting coefficients for the three metrics may need tuning for specific deployment scenarios. Future work is suggested in extending the approach to heterogeneous MANETs, developing adaptive weight selection mechanisms, and validating the algorithm on real‑world testbeds.