A Distributed Data Collection Algorithm for Wireless Sensor Networks with Persistent Storage Nodes

A distributed data collection algorithm to accurately store and forward information obtained by wireless sensor networks is proposed. The proposed algorithm does not depend on the sensor network topology, routing tables, or geographic locations of sensor nodes, but rather makes use of uniformly distributed storage nodes. Analytical and simulation results for this algorithm show that, with high probability, the data disseminated by the sensor nodes can be precisely collected by querying any small set of storage nodes.

💡 Research Summary

The paper addresses the problem of reliable data collection in large‑scale wireless sensor networks (WSNs) where sensor nodes have limited memory, processing power, and battery life, and may disappear abruptly due to failures or energy depletion. Existing approaches often rely on knowledge of the network topology, routing tables, or geographic coordinates, which limits their applicability in hostile or inaccessible environments. To overcome these constraints, the authors propose a fully distributed data collection algorithm, named DSA‑I, that exploits a set of uniformly distributed “persistent storage nodes” (also called cluster heads) without requiring any routing or location information.

Network model and assumptions

• The deployment region R is a square of side L (e.g., L = 100).
• k sensor nodes (≈ 80 % of all nodes) and n − k storage nodes (≈ 10‑20 % of all nodes) are placed independently and uniformly at random over R.
• Sensor nodes have homogeneous capabilities (limited memory, bandwidth, and power) and cannot maintain routing tables or geographic tables.
• Each storage node possesses a large memory buffer of size M, which is divided into ε = ⌊M/c⌋ smaller slots of size c.
• Nodes can discover their immediate neighbors by broadcasting a simple flooding query; a storage node can multicast to all its neighbors.
• Packets consist of (node ID, sensed data x, flag) where flag = 0 denotes an initialization packet and flag = 1 denotes an update packet.

Algorithm DSA‑I

1. Clustering phase – Every storage node broadcasts a beacon containing its ID. Sensor nodes that receive the beacon store the ID of that storage node; thus each sensor builds a list of storage nodes within its radio range δ.
2. Sensing phase – Each sensor collects environmental data, forms a packet (ID, x, flag), and multicasts it to all storage nodes in its list.
3. Data collection and storage phase – Upon receiving a packet, a storage node stores the data in an empty buffer if flag = 0 (initialization). If flag = 1 (update), the node updates its stored value by XORing the new data with the existing content (y ← y ⊕ x). This simple linear operation enables the storage node to aggregate multiple sensor readings without keeping each packet separately.
4. Querying phase – A base station (or data collector) queries a subset of the storage nodes. Let h be the number of queried storage nodes and η = h/(n − k) the query ratio. The base station receives the current contents of the queried buffers and attempts to reconstruct all k original sensor packets using linear decoding (e.g., Gaussian elimination).

Theoretical analysis

• Lemma 1 states that if each storage node has at least ε ≥ k/(n − k) buffer slots, then with high probability the collector can recover all sensor data. This condition guarantees that, on average, each storage node can accommodate at least one distinct sensor’s packet.
• Lemma 2 derives the probability that a given sensor lies within the radio range δ of a particular storage node: P = πδ²/L² − a/L², where a accounts for the portion of the storage node’s coverage that falls outside the deployment region.
• Lemma 3 extends this to all storage nodes, yielding the probability that a sensor is covered by at least one storage node: (πδ²/L² − a/L²)·(n − k).
• Lemma 4 gives the probability that all sensors fall within the range of a single storage node, which is useful for analyzing worst‑case concentration of data.

These lemmas collectively show that, provided the radio range and buffer size are chosen appropriately, the random placement of storage nodes yields sufficient coverage and redundancy for successful data recovery without any explicit routing.

Simulation study
The authors evaluate DSA‑I through extensive simulations on a square region of side L = 100. Key parameters varied include the total number of nodes (n = 250 – 1750), buffer capacity per storage node (ε = 20, 40), and radio range δ (0.5 – 2 distance units). Performance metrics are:

• Successful decoding probability (Pₛ) – the probability that all k source packets are recovered.
• Query ratio (η) – fraction of storage nodes queried.
• Revealed sensors ratio (ρ) – fraction of sensors whose data is successfully retrieved.

Results indicate that:

1. As n grows, Pₛ improves because more storage nodes increase the likelihood that each sensor is covered.
2. For a fixed buffer size, querying roughly 20 %–30 % of the storage nodes (η ≈ 0.2‑0.3) is sufficient to achieve ρ ≈ 1, i.e., almost all sensor data is recovered.
3. Larger buffers (ε = 40) allow the system to tolerate smaller η while maintaining high Pₛ.
4. The radio range δ exhibits a sweet spot: too small a range leaves many sensors uncovered, reducing Pₛ; too large a range creates excessive overlap, leading to many XORed packets that become difficult to decode, again lowering Pₛ.
5. In sparse deployments (e.g., n = 250) with a small δ, the system can recover at most about 60 % of the data, highlighting the importance of adequate coverage.

Related work
The paper contrasts DSA‑I with fountain‑code based schemes that require geographic routing and knowledge of node positions (e.g., Dimakis et al.) and with prior persistence‑oriented algorithms that store data directly on sensors using random walks. Unlike those, DSA‑I places the storage burden on dedicated storage nodes, eliminates the need for routing tables, and automatically adapts to node failures because the clustering is formed dynamically via beacon flooding.

Conclusions and future directions
DSA‑I demonstrates that a WSN can achieve reliable, persistent data collection without any routing or location information, simply by leveraging uniformly distributed storage nodes that act as cluster heads and perform lightweight XOR aggregation. The algorithm scales well with network size, tolerates node failures, and requires only a modest fraction of storage nodes to be queried for full data recovery. Future research avenues include (i) extending the model to non‑uniform or mobile node distributions, (ii) implementing the protocol on real sensor hardware to assess energy consumption and latency, and (iii) integrating security mechanisms (e.g., encryption, authentication) to protect the aggregated data against adversarial attacks.