💻 computer science

Adaptive Quantum-Inspired Swarm Learning (AQISL) for Energy-Efficient Cluster Head Selection in IoT Infrastructure

This paper proposes the Adaptive Quantum-Inspired Swarm Learning (AQISL) algorithm to optimize cluster head selection in large-scale IoT networks, effectively reducing energy consumption, packet loss, and transmission latency while enhancing load balancing and overall network stability.

Original authors: C. Bala Subramanian, R. Ramkumar, S Shanmugapriya

Published 2026-10-07

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: C. Bala Subramanian, R. Ramkumar, S Shanmugapriya

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a vast, silent landscape dotted with thousands of tiny, battery-powered devices. These are the sensors of the Internet of Things, quietly monitoring everything from the humidity in a greenhouse to the structural integrity of a bridge. They are the eyes and ears of a modern world, yet they face a fundamental limitation: they run on finite energy. Once their batteries die, they go silent, and the network they form begins to crumble. In many real-world scenarios, such as remote environmental monitoring or disaster zones, replacing these batteries is impossible. The challenge, then, is not just making these devices talk to each other, but ensuring they do so without exhausting their power too quickly. To solve this, engineers often group these devices into small neighborhoods called clusters. Within each neighborhood, one device is chosen to act as a leader, or "cluster head." This leader gathers information from its neighbors, combines it into a single message, and sends it to a central base station. This method saves energy because the ordinary devices only need to whisper their data to a nearby leader, rather than shouting it across the entire network to the base station. However, choosing the right leader is a difficult puzzle. If the same device is chosen too often, it burns out fast, leaving a gap in the network. If the wrong device is chosen, the message might get lost or take too long to arrive.

Researchers C. Bala Subramanian, R. Ramkumar, and Shanmugapriya S. from the Kalasalingam Academy of Research and Education in India have proposed a new way to solve this puzzle. They developed a system called Adaptive Quantum-Inspired Swarm Learning, or AQISL, designed to intelligently select the best leader for each cluster at any given moment. Their approach combines two distinct ways of thinking about problem-solving. The first draws from the principles of quantum mechanics, not to build a quantum computer, but to use a mathematical style that allows for many possibilities to exist at once. This helps the system explore a wide range of potential solutions without getting stuck on a single, suboptimal choice. The second part mimics the behavior of a swarm, like a flock of birds or a colony of ants, where individual members learn from their own experiences and from the collective success of the group. By blending these two ideas, the algorithm can adapt to changing conditions in real-time, such as when a device's battery runs low or when the network topology shifts.

The researchers tested their idea using a computer simulation that modeled a large network of sensor nodes scattered across a square area. In this virtual environment, the devices communicated, consumed energy, and died off just as they would in the real world. The team compared their new AQISL method against several established techniques, including older random selection methods and other advanced optimization algorithms. The results showed a clear advantage for the new approach. In the simulations, the network using AQISL lasted significantly longer than those using traditional methods. Specifically, the first device in the AQISL network survived for 890 rounds of communication, whereas the first device in a standard LEACH network died after only 520 rounds. When half of the devices in the network had run out of power, the AQISL system had reached 1,320 rounds, compared to 780 rounds for the standard method. This extended lifespan suggests that the algorithm is much better at distributing the workload, ensuring that no single device is overworked while others sit idle.

Beyond just keeping the network alive, the new method improved the quality of the communication itself. The simulations indicated that the AQISL system reduced the number of lost messages to 7.9 percent, a significant improvement over the 18.5 percent loss rate seen in the standard LEACH method. The time it took for data to travel from a sensor to the base station, known as delay, was also reduced to an average of 102 milliseconds, which is faster than the 160 milliseconds recorded for other leading methods. The energy consumed by the network in each round of communication dropped to 0.52 Joules, a figure lower than the 0.74 Joules required by other sophisticated algorithms. Perhaps most importantly, the system maintained a high degree of balance. The researchers measured a "load balancing index" of 0.91 for their method, indicating that the energy drain was spread very evenly across the network, preventing the formation of "energy holes" where clusters of dead devices would leave the network blind.

The study suggests that this hybrid approach offers a robust solution for the future of large-scale sensor networks. While the method did require slightly more control data to be sent between devices to coordinate the selection process, the trade-off was well worth it for the gains in stability and efficiency. The researchers note that the algorithm's ability to adapt to dynamic conditions makes it particularly suitable for environments where devices might move or fail unexpectedly. By simulating these complex interactions, the team demonstrated that it is possible to create a network that is not only longer-lasting but also more reliable and efficient. The findings point toward a future where the Internet of Things can operate more autonomously, managing its own resources with a level of intelligence that mimics the collective wisdom of nature, ensuring that the vast web of connected devices continues to function long after the initial batteries would have otherwise failed.

Technical Summary: Adaptive Quantum-Inspired Swarm Learning (AQISL) for Energy-Efficient Cluster Head Selection in IoT Infrastructure

1. Problem Statement

The construction of large-scale Internet of Things (IoT) networks faces significant challenges regarding communication efficiency and energy conservation. Most IoT nodes are battery-powered and often deployed in remote areas where recharging or replacement is impractical. In these networks, data transmission is the most energy-intensive operation. While clustering is a standard approach to organize nodes hierarchically and reduce communication overhead by aggregating data at a Cluster Head (CH), selecting the optimal CH remains a complex problem.

Existing methods suffer from several limitations:

Traditional Protocols (e.g., LEACH, HEED): Often rely on random or static parameter-based selection, leading to unbalanced energy consumption, premature node death, and suboptimal network lifespans.
Metaheuristic Algorithms (e.g., GA, PSO, ABC): While they offer global search capabilities, they frequently suffer from premature convergence (getting stuck in local optima), high computational complexity, and slow convergence times, making them less suitable for dynamic, large-scale IoT environments.
Machine Learning Approaches: While adaptive, they often incur high computational overhead and memory requirements that exceed the capabilities of resource-constrained sensor nodes.

The core challenge is to develop a CH selection mechanism that balances exploration and exploitation, adapts to dynamic network conditions (energy levels, topology changes), and minimizes energy consumption without incurring excessive computational costs.

2. Methodology: Adaptive Quantum-Inspired Swarm Learning (AQISL)

The paper proposes the Adaptive Quantum-Inspired Swarm Learning (AQISL) algorithm, a hybrid approach that integrates quantum-inspired probabilistic modeling with swarm intelligence principles to optimize CH selection.

2.1 Core Components

Quantum-Inspired Representation:
- Each node is represented by a quantum probability model, allowing it to exist in multiple states simultaneously. This is defined by probability amplitudes ( $\alpha$ and $\beta$ ) where $|\alpha|^2 + |\beta|^2 = 1$ .
- The state is updated using a rotation gate mechanism (Equations 6 and 7) to evolve probabilities over iterations. This mechanism enhances solution diversity and mitigates the risk of premature convergence common in traditional metaheuristics.
- A node is selected as a CH if its probability $P_i$ exceeds a dynamic threshold $\tau$ .
Swarm Learning Strategy:
- The algorithm employs a Particle Swarm Optimization (PSO)-like mechanism where candidate solutions update their positions and velocities based on individual best ( $P_{best}$ ) and global best ( $G_{best}$ ) experiences.
- An adaptive inertia weight ( $\omega$ ) is used to balance exploration (early stages) and exploitation (later stages).
- The swarm learns from collective experiences to converge faster toward optimal CH configurations.
Fitness Function Design:
- The algorithm evaluates candidate CH configurations using a composite fitness function ( $F_i$ $F_{i}$ ) that integrates:
  1. Residual Energy ( $E_{res}$ ): Prioritizes nodes with higher remaining energy.
  2. Distance to Base Station ( $d_{bs}$ ): Favors nodes closer to the sink to reduce transmission energy.
  3. Node Density ( $\rho$ ): Considers the number of nodes in the vicinity to ensure efficient data aggregation.
- Transmission and reception energy models are incorporated to calculate total communication costs.

2.2 Algorithm Workflow

The AQISL process involves 13 steps, ranging from network configuration and candidate population generation to iterative updates of quantum states and swarm positions. The algorithm dynamically recalculates fitness values, updates selection probabilities, and selects the optimal set of CHs ( $CH_{opt}$ ) that maximizes the fitness function while ensuring load balancing.

3. Key Contributions

The study identifies three primary contributions:

Novel Hybrid Algorithm: The development of AQISL, which uniquely combines quantum-inspired probabilistic modeling with swarm learning to achieve a superior balance between global exploration and local exploitation.
Integrated Fitness Function: The creation of a unified fitness function that simultaneously optimizes residual energy, distance to the base station, and node density, addressing multiple constraints in a single optimization step.
Adaptive Mechanism: The implementation of an adaptive system capable of responding to dynamic network conditions, such as fluctuating energy levels and changing topologies, without requiring extensive retraining or manual parameter tuning.

4. Experimental Results

The proposed method was evaluated using MATLAB simulations in a 2D square area with randomly distributed sensor nodes. The performance of AQISL was compared against traditional and metaheuristic-based methods, including LEACH, HEED, PSO-based, ABC-based, and BCO-based CH selection.

Key Performance Metrics:

Network Lifetime: AQISL significantly extended the network lifespan. The "First Node Death" occurred at 890 rounds (compared to 520 for LEACH and 760 for BCO), and "Half Node Death" occurred at 1320 rounds (compared to 780 for LEACH and 1140 for BCO).
Energy Efficiency: The energy consumption per round was reduced to 0.52 J, outperforming PSO (0.74 J) and ABC (0.69 J).
Reliability: Packet loss was minimized to 7.9%, a significant improvement over LEACH (18.5%) and BCO (10.4%).
Latency: End-to-end delay was reduced to 102 ms, compared to 160 ms for HEED and 142 ms for PSO.
Load Balancing: The Load Balancing Index improved to 0.91, indicating a more even distribution of energy consumption across nodes.
Control Overhead: While AQISL had a slightly higher control overhead (4900 bits) compared to LEACH (4200 bits), it remained lower than other metaheuristic approaches like BCO (5700 bits) and ABC (5400 bits).

5. Significance and Claims

The paper claims that AQISL offers a stable, energy-efficient, and reliable solution for CH selection in large-scale, dynamic IoT environments. By effectively preventing premature convergence and balancing the trade-off between exploration and exploitation, the algorithm ensures:

Extended Network Lifetime: Through optimized energy distribution and reduced communication costs.
Enhanced Robustness: By adapting to dynamic changes in node energy and network topology.
Improved Quality of Service: Evidenced by lower packet loss, reduced latency, and higher throughput.

The authors conclude that while the control overhead is marginally higher than simple probabilistic methods like LEACH, the substantial gains in network longevity, energy efficiency, and reliability validate AQISL as a superior approach for future IoT infrastructure management. The study emphasizes that the method is scalable and suitable for heterogeneous device environments where energy conservation is critical.