In recent years, the increasing frequency of natural disasters such as earthquakes, wildfires, and hurricanes has highlighted a critical challenge: the swift destruction of local communication infrastructure. This breakdown severs communication links between disaster zones and the outside world, creating severe obstacles for rescue and relief operations. In this context, emergency communication systems are not just beneficial but vital. Unmanned Aerial Vehicles (UAVs), or drones, have emerged as pivotal assets due to their simple operation, high mobility, and strong collaborative capabilities. Their applications have expanded significantly, encompassing tasks like disaster reconnaissance, supply delivery, firefighting support, and the rapid deployment of temporary communication bases. A formation of multiple small drones, orchestrated by a central command, operates with high data rates and dynamic adaptability, making it a prime candidate for emergency communication scenarios.
Inspired by the collective behavior of insects like bees, a drone formation often operates in a clustered manner to execute specific missions efficiently. This clustered architecture enhances inter-drone communication, expands the network’s collective sensing and data acquisition capabilities over a disaster area, and ultimately improves rescue efficiency. However, the proliferation of drone technology introduces new challenges for formation communication, including scarce radio frequency spectrum, mutual interference, and network connectivity issues. The sub-6G spectrum is already heavily utilized by terrestrial cellular networks. The widespread operation of drone formations can severely disrupt these ground networks, causing a system capacity drop far beyond planned expectations and degrading communication quality for ground users. Furthermore, during critical missions, the connectivity of a drone swarm can be compromised by complex atmospheric conditions and strong electromagnetic interference, leading to degraded system performance and reduced mission efficiency.
Wireless Ultraviolet (UV) communication presents a compelling solution to these challenges. Operating in the solar-blind spectrum (200–280 nm), UV light is heavily absorbed and scattered by the atmosphere. This characteristic grants UV-based communication networks distinct advantages: low probability of intercept and detection, non-line-of-sight (NLOS) capability through atmospheric scattering, omnidirectionality, and all-weather operation. These properties make UV communication ideal for establishing secure, robust, short-range mobile local area networks. When integrated as a complementary technology to traditional wireless systems on drone platforms, it can significantly enhance the endurance, reconnaissance, and collaborative rescue capabilities of a drone formation. The formation typically functions in a “leader-cluster head-member” mode, using wireless UV links to form a self-organizing, multi-hop network for collecting, processing, and exchanging information about the monitored area, thereby facilitating task allocation and coordination.
During actual disaster relief missions, the energy carried by a drone formation is inherently limited. This constraint is exacerbated in UV communication due to the high path loss of the UV channel. Consequently, designing highly efficient clustering protocols and energy-optimized routing algorithms is of paramount importance for extending the operational lifetime of the drone formation and maximizing its rescue potential. The classic Low Energy Adaptive Clustering Hierarchy (LEACH) protocol is a foundational clustering algorithm where nodes have an equal probability of becoming a cluster head (CH) per round. However, it does not account for the residual energy of nodes or their geographical distribution during CH election. This oversight can lead to distant CH nodes depleting their energy prematurely due to long-distance transmission to the leader, causing early node death and a significant reduction in overall network lifetime. To address energy concerns, improvements have been proposed, such as modifying the CH election threshold based on residual energy or using a single relay node for inter-cluster communication. Yet, opportunities remain for more comprehensive optimization of both the clustering process and the multi-hop routing between CHs and the formation leader.
To substantially prolong the effective mission time of UV communication-enabled drone formations, this work proposes a novel routing optimization algorithm named Rcomp JAYA LEACH (RJLEACH). This algorithm enhances the classic LEACH protocol in two key phases. First, it introduces a competition radius mechanism during the CH election phase to ensure a more balanced geographical distribution of CHs and reduce variance in residual energy across the drone formation. Second, for inter-cluster communication, it employs a Discrete JAYA (DJAYA) intelligent optimization algorithm to dynamically search for the optimal multi-hop routing path from a CH to the formation leader, considering both path distance and the residual energy of relay nodes. By integrating these improvements, the RJLEACH algorithm aims to balance energy consumption across the network and minimize total transmission energy, thereby extending the functional lifespan of the entire drone formation and buying precious time for critical missions like disaster relief and emergency communication.
System Model for UV-Enabled Drone Formation
The routing optimization problem is framed within a clustered network model for the drone formation. As illustrated in the accompanying figure, the entire network of UAVs is organized into several clusters. Each cluster has one drone elected as the Cluster Head (CH). The CH acts as a local coordinator, aggregating data from its member drones and communicating with the formation leader, thereby alleviating the leader’s control and communication burden. The formation leader, equipped with high-performance computing units, serves as the core command node, making mission-critical decisions and managing overall task allocation.
The operational model works in rounds. Within each round:
1. Cluster heads are elected from the drone formation.
2. Non-CH drones (members) join the nearest CH to form clusters.
3. Member drones send their status data to their respective CH.
4. The CH performs data fusion to eliminate redundancy.
5. The CH transmits the aggregated data to the formation leader, either directly or via a multi-hop path through other CHs acting as relays.
6. The leader sends back control signals.
At the end of the round, clusters dissolve, and a new round begins with a fresh CH election. This periodic re-clustering helps balance energy load across the drone formation.
For communication, we focus on the Non-Line-of-Sight (NLOS) UV channel, which is essential for maintaining links in obstructed disaster environments where direct alignment is impossible. In NLOS UV communication, transmitted photons are scattered by atmospheric particles, allowing reception as long as the transmitter and receiver fields of view overlap. We consider a directed NLOS link geometry. The path loss \(L\), defined as the ratio of transmitted power \(E_T\) to received power \(E_R\), is a critical metric. For relatively short ranges (r < 1 km) typical of drone formation operations, the complex path loss model can be simplified to a closed-form expression:
$$L = \xi r^{\alpha}$$
where \(\xi\) is the path loss factor and \(\alpha\) is the path loss exponent. Both parameters depend on the transmitter and receiver elevation angles (\(\theta_1, \theta_2\)). For our simulations, considering a balance between link performance and practical drone formation flight, we set \(\theta_1 = \theta_2 = 40^\circ\), where path loss variation is relatively stable.
Based on this, the energy lost due to transmission attenuation \(L\) for sending one bit is:
$$E_L = E_T (1 – 1/L)$$
Therefore, the energy consumed to transmit \(k\) bits over a distance \(r\) is:
$$E_{TX}(k, r) = k (E_T + E_L) = k E_T (2 – 1/L)$$
The energy to receive \(k\) bits is:
$$E_{RX}(k) = k E_R$$
where \(E_R\) is the receiver electronics energy per bit. A CH also consumes energy for data fusion. Fusing reports from \(m\) member nodes for \(k\)-bit data packets requires:
$$E_{C}(m, k) = (m+1) k E_{DA}$$
where \(E_{DA}\) is the energy cost per bit for data aggregation.
The key simulation parameters for modeling the UV communication drone formation are summarized in the table below.
| Parameter | Value |
|---|---|
| UV Wavelength \(\lambda\) | 266 nm |
| Initial Node Energy \(E_0\) | 300 J |
| Transmit/Receive Energy per bit \(E_T, E_R\) | 8 mJ |
| Data Aggregation Energy per bit \(E_{DA}\) | 0.8 mJ |
| Tx & Rx Elevation Angles \(\theta_1, \theta_2\) | 40° |
| Path Loss Factor \(\xi\) | 1.69 × 10⁸ |
| Path Loss Exponent \(\alpha\) | 1.3498 |
| Data Packet Size (per round) | 5000 bits |
| Control Packet Size (per round) | 400 bits |
Design of the RJLEACH Routing Optimization Algorithm
The proposed RJLEACH algorithm enhances the drone formation’s network lifetime through targeted improvements in the CH election and inter-cluster routing phases.
Cluster Formation with Competition Radius
The first improvement addresses the CH election. In classic LEACH, each node \(i\) becomes a candidate CH in a round if a random number \(r_0\) is less than a threshold \(T(n)\):
$$T(n) = \begin{cases} \frac{p}{1 – p \left( r \mod \frac{1}{p} \right)} & \text{if } n \in G \\ 0 & \text{otherwise} \end{cases}$$
where \(p\) is the desired CH percentage, \(r\) is the round number, and \(G\) is the set of nodes that haven’t been CH in the last \(1/p\) rounds.
RJLEACH modifies this by introducing a competition radius \(R_i\) for each candidate CH. This radius is not fixed but is adaptively calculated based on the candidate’s distance to the formation leader \(d_{i,leader}\):
$$R_i = \left(1 – r_c \frac{d_{max} – d_{i,leader}}{d_{max} – d_{min}}\right) R_0$$
Here, \(d_{max}\) and \(d_{min}\) are the maximum and minimum distances from any node to the leader in the drone formation, \(r_c\) is a control parameter in [0,1], and \(R_0\) is the maximum competition radius (empirically set to half the formation’s dimensions). This formula ensures that nodes farther from the leader have a smaller competition radius, promoting the election of more CHs in distant areas to shorten their direct communication hops later.
After generating this radius, each candidate CH node \(s_i\) identifies its “neighbor candidate CHs” within a distance defined by \(\max(R_i, R_j)\). Within this neighbor set, only the node with the highest residual energy announces itself as the final CH for this round. Other candidates in the set withdraw from the election. This mechanism, governed by the competition radius, results in CHs that are well-distributed geographically and possess high residual energy, effectively balancing the initial load across the drone formation.
Inter-Cluster Routing Optimization with Discrete JAYA (DJAYA)
The second improvement targets the energy-intensive communication from CHs to the leader. A simple direct transmission can quickly drain a distant CH’s battery. Therefore, we introduce a distance threshold \(D_{max}\). If a CH’s distance to the leader is less than \(D_{max}\), it transmits directly. Otherwise, it must find a multi-hop path through other CHs acting as relay nodes.
The core of RJLEACH is to find the optimal relay path. For a CH \(n_i\) needing \(k\) relay nodes to reach the leader \(C\), a path \(P\) can be represented as \((n_i, R_1, R_2, …, R_k, C)\). We define a fitness function \(F(P)\) to evaluate any candidate path, aiming to minimize both the total transmission distance and the inverse of the total residual energy of the relay nodes (to avoid low-energy nodes):
$$F(P) = \sum_{j=1}^{k+1} d_{j, j+1} + \left( – \sum_{j=1}^{k} E_{rest}(R_j) \right) = Distance(P) – Energy(P)$$
Minimizing \(F(P)\) corresponds to finding a short path using relays with high residual energy.
We employ the Discrete JAYA (DJAYA) algorithm, a parameter-less, population-based metaheuristic, to solve this optimization problem for each CH requiring a multi-hop path. Each candidate path is a “seed” (solution) in the population. The algorithm iteratively improves the population by moving seeds away from the worst-performing seed and toward the best-performing seed, with a stochastic component. Search trend parameters (\(PST1, PST2\)) are used to balance exploration and exploitation in the solution space. The DJAYA process for the drone formation routing is as follows:
- Initialize: Set population size \(N\), max iterations \(I_{max}\), \(PST1\), \(PST2\), relay count \(k\). Generate an initial population of random paths for a given source CH.
- Evaluate: Calculate fitness \(F(P)\) for all seeds. Identify the best seed \(P_{best}\) and worst seed \(P_{worst}\).
- Generate Candidate: For each seed \(P_{current}\), generate a candidate seed \(P_{candidate}\) by modifying the path based on \(P_{current}\), \(P_{best}\), \(P_{worst}\), \(PST1\), and \(PST2\). This often involves randomly swapping or replacing relay nodes from the current path with other CHs from the network.
- Select: If \(F(P_{candidate}) < F(P_{current})\), replace \(P_{current}\) with \(P_{candidate}\).
- Terminate: Repeat steps 2-4 until \(I_{max}\) is reached. The best seed in the final population is selected as the optimal routing path for that CH.
This DJAYA-based routing is executed per round for all CHs located beyond \(D_{max}\), dynamically adapting to the changing residual energy landscape of the drone formation.
Simulation Results and Comparative Analysis
To validate the performance of the RJLEACH algorithm, we simulated a drone formation of 100 nodes randomly deployed in a 200m × 200m area, with the leader at the center. Key DJAYA parameters were set as: \(PST1 = PST2 = 0.5\), population size \(N=50\), iterations \(I_{max}=50\). The CH probability \(p\) was 20%, competition radius parameters \(r_c=0.5, R_0=90m\). The optimal number of relay nodes \(k\) was found to be 2, and the distance threshold \(D_{max}\) was set to 100m through preliminary parameter analysis. We compare RJLEACH against the classic LEACH protocol and two intermediate variants: LEACH with only the Competition Radius module (LEACH+Rcomp) and LEACH with only the DJAYA routing module (LEACH+DJAYA).
The impact of node density was also studied by scaling the deployment area. The results, summarized below, confirm that lower density (larger area) leads to faster energy depletion due to longer average communication distances, underscoring the need for efficient routing in sparse drone formations.
| Deployment Area | Avg. Energy Depletion Round | First Node Death Round | vs. 200m² Baseline |
|---|---|---|---|
| 200m × 200m | ~1000 | 974 | Baseline |
| 300m × 300m | ~800 | 800 | 17.9% earlier |
| 400m × 400m | ~700 | 620 | 36.3% earlier |
The effectiveness of the competition radius mechanism is evident in the CH distribution. While classic LEACH produces randomly distributed, often uneven clusters with some isolated CHs, RJLEACH ensures well-spaced, stable CHs throughout the operational lifetime of the drone formation. The average number of CHs per round in LEACH fluctuates significantly (between 12 and 26). In contrast, RJLEACH maintains a stable average of about 7 CHs with minimal fluctuation (between 4 and 11), leading to more balanced cluster sizes.
A critical metric for the drone formation’s health is the variance of residual energy among nodes. High variance indicates unbalanced energy consumption, leading to early node failures. The following table compares the average residual energy variance during the network’s stable lifetime for the four algorithms:
| Algorithm | Average Residual Energy Variance | Reduction vs. LEACH |
|---|---|---|
| LEACH (Baseline) | 130.29 | — |
| LEACH + DJAYA | 90.62 | 30.4% |
| LEACH + Rcomp | 32.11 | 75.3% |
| RJLEACH (Proposed) | 29.46 | 77.4% |
The proposed RJLEACH algorithm achieves the lowest energy variance, demonstrating its superior capability in balancing energy consumption across the entire drone formation. This directly translates to a prolonged network lifetime. The key lifetime metrics are compared below:
| Lifetime Metric | LEACH | LEACH+DJAYA | LEACH+Rcomp | RJLEACH |
|---|---|---|---|---|
| First Node Death (FND) | 740 | 760 | 915 | 975 |
| Half of the Nodes Dead (HND) | 870 | 910 | 921 | 990 |
| Last Node Death (LND) | 1000 | 1019 | 950 | 1041 |
The results are striking. Compared to the classic LEACH protocol, the RJLEACH algorithm extends the time until the first drone in the formation depletes its energy (FND) by 31.8% (from 740 to 975 rounds). More importantly, it extends the time until half the drones are dead (HND) by 13.8% (from 870 to 990 rounds). The HND metric is particularly crucial as it marks the point where the drone formation’s coverage and data collection capabilities are severely degraded. RJLEACH significantly delays this point, ensuring a longer period of high-quality, cooperative operation for the drone formation during critical emergency missions.
Conclusion
This work addresses the vital challenge of energy-efficient communication in drone formations deployed for emergency scenarios. By integrating the advantages of secure, NLOS ultraviolet light communication with an intelligent clustering and routing protocol, we have developed the RJLEACH algorithm. RJLEACH enhances the standard LEACH protocol through a competition-radius-based CH election that promotes balanced node energy and a DJAYA-optimized multi-hop routing that minimizes overall transmission energy consumption.
Simulation results demonstrate that the proposed algorithm offers substantial improvements over the classic LEACH protocol for managing a drone formation. RJLEACH reduces the residual energy variance among drones by over 77%, ensuring balanced energy consumption. Most significantly, it extends the functional operational lifetime of the formation, delaying the death of the first node by 31.8% and the death of half the nodes by 13.8%. This extension represents valuable additional time for a drone formation to execute critical sensing, reconnaissance, and communication relay tasks in disaster-stricken areas where every moment counts. The RJLEACH algorithm thus presents a robust and efficient solution for enhancing the resilience and endurance of UV-based cooperative drone networks in emergency response applications.