Hybrid Optimization Algorithm of OAFS and ICS with CNN-based Energy Level Prediction for Wireless Sensor Networks

S. Jagadeesh, C. Mahesh, Amit Kumar, N. Noor AlleemaResearch Square 2023 · DOI

TL;DR

This paper introduces a hybrid optimization framework, OAFS-ICS, designed to extend the lifespan of Wireless Sensor Networks (WSNs). By combining deep learning (DCNN) for energy prediction with advanced metaheuristic algorithms for clustering and routing, the authors achieved significantly higher data throughput and longer network survival compared to traditional optimization methods.

What Problem Does This Paper Address?

Wireless Sensor Networks (WSNs) are the backbone of the Internet of Things (IoT), consisting of hundreds of tiny, battery-powered nodes that monitor environments. The "Achilles' heel" of these networks is energy: once a node runs out of power, it’s usually dead for good.

To save energy, nodes are organized into clusters, where a Cluster Head (CH) aggregates data from its neighbors and sends it to a Base Station (BS). However, choosing the right CH and finding the most efficient path to the BS (routing) are "NP-hard" problems—mathematically complex tasks where finding the absolute best solution is computationally exhausting. Most existing methods struggle with "hotspots" (nodes dying early due to heavy traffic) or inefficient pathfinding, leading to premature network failure.

Methods

The authors propose a three-pronged "hybrid" strategy to tackle these challenges:

  1. Energy Prediction via DCNN: Before selecting Cluster Heads, the system uses a refined Deep Convolutional Neural Network (DCNN) to predict the future energy levels of nodes. This allows the network to proactively avoid choosing nodes that are about to run out of juice.
  2. Clustering with OAFS: For grouping nodes, the paper uses Oppositional Artificial Fish Swarm (OAFS). This is an improvement on the standard Fish Swarm algorithm that incorporates "Opposition-Based Learning" (OBL). OBL helps the algorithm explore the search space more effectively by considering both a candidate solution and its opposite, preventing the system from getting stuck in "local optima" (mediocre solutions).
  3. Routing with ICS: For the actual data transmission, the authors developed Improved Cuckoo Search (ICS). By integrating Simulated Annealing and orthogonal design into the standard Cuckoo Search, the algorithm finds more efficient, multi-hop paths from the CHs to the Base Station.

The selection of CHs is governed by a Fitness Function that balances four critical factors: residual energy, node degree (how many neighbors it has), distance to the Base Station, and node centrality.

Datasets

The study relies on simulated environments created in MATLAB rather than static datasets. The authors deployed a network of 300 nodes randomly distributed across a 200m x 200m area.

They tested the model across three distinct geographical scenarios:

  • Scenario 1 (S1): Base Station located at the dead center.
  • Scenario 2 (S2): Base Station placed in a secluded corner.
  • Scenario 3 (S3): Base Station located in an "inappropriate" (non-optimal) location.

Key Insights and Ideas

  • Proactive vs. Reactive: Most WSN protocols look at current energy. By using a DCNN to predict energy levels, this model acts like a chess player looking several moves ahead, ensuring the network structure remains stable for longer.
  • The Power of Hybridization: No single algorithm is perfect. By using OAFS for the "macro" task of clustering and ICS for the "micro" task of routing, the authors play to the strengths of different metaheuristics.
  • Resilience to BS Location: One of the most interesting takeaways is the model's stability. Even when the Base Station was moved to an "inappropriate" location (Scenario 3), the OAFS-ICS model maintained a significantly longer lifespan than its competitors.

Key Results

The OAFS-ICS model was compared against several industry-standard benchmarks, including Artificial Bee Colony (ABC) and Cuckoo Search (CSA).

  • Network Lifetime: In Scenario 1, the First Node Die (FND)—a critical metric for network reliability—occurred at 1,897 rounds for OAFS-ICS, nearly double the 1,040 rounds seen in the standard CSA model.
  • Data Throughput: The proposed model successfully delivered 19,643 packets to the Base Station in Scenario 1, outperforming all other tested algorithms.
  • Energy Consumption: OAFS-ICS reached a state of 1 Joule consumption at 420 rounds, whereas the KHOA and CSA methods hit that mark much earlier (192 and 147 rounds, respectively), proving the hybrid model is significantly more "frugal" with power.

Limitations and Future Work

While the results are impressive, the paper notes a few areas for improvement:

  • Hotspot Issues: While the model extends life, it doesn't entirely eliminate the "hotspot" problem where nodes near the Base Station die faster due to relaying more traffic. The authors suggest "uneven clustering" as a future solution.
  • Simulation vs. Reality: The study is entirely simulation-based. Real-world factors like physical obstacles, weather interference, and hardware-specific battery discharge curves were not discussed and would be necessary for commercial deployment.
  • Computational Overhead: The paper does not deeply analyze the computational cost of running a DCNN and two metaheuristic optimizations on what are typically very low-power sensor processors.

Summary updated

Abstract

<title>Abstract</title> In order to maximize network duration and attain power efficiency in wireless sensor networks, clustering, and routing are two notable optimization problems (WSNs). The clustering and routing procedure is an example of an NP-hard issue that can be solved using a metaheuristic optimization method. Clustering is a suitable procedure that is frequently used to improve network power efficiency. Concurrently, difficult cluster head (CH) election and potential Base Station (BS) pathways increase energy consumption and cut down the lifetime of the WSN. This paper proposes an Improved Cuckoo Search (ICS) routing method along with Oppositional Artificial Fish Swarm (OAFS) based clustering as a solution to this issue. The OAFS-ICS approach that has been suggested makes good use of OAFS-based clustering to choose the CHs. In this case, a refined Deep Convolutional Neural Network (DCNN) is worn to make predictions about the energy level. The CH parameters like distance to BS (DBS), residual energy, node degree (ND) and node centrality are used to calculate a fitness function (FF). Several scenarios are utilised to calculate the performance of the current technique depending on the number of nodes. Numerous simulations were conducted in order to confirm the supplied model's superiority. The simulation results demonstrated that the OAFS-ICS technique beat the comparison methods in terms of a variety of criteria.
BibTeX
@inproceedings{jagadeesh2023hybrid,
  author = {S. Jagadeesh and C. Mahesh and Amit Kumar and N. Noor Alleema},
  title = {{Hybrid Optimization Algorithm of OAFS and ICS with CNN-based Energy Level Prediction for Wireless Sensor Networks}},
  booktitle = {Research Square},
  year = {2023},
  doi = {10.21203/rs.3.rs-3676799/v1},
}
Figure