In large-scale Internet of Things deployments, wireless sensor networks (WSN) are frequently partitioned into multiple geographically isolated sub-regions due to terrain, obstacles, or deployment constraints. This spatial segmentation poses significant challenges for traditional multi-hop data collection mechanisms. Unmanned aerial vehicles (UAVs), especially UAV drones, offer a flexible and mobile solution to overcome these obstacles. However, existing UAV drone-assisted data collection schemes often suffer from excessive path redundancy and energy waste because each UAV drone works independently and must return to the base station to deliver collected data. To address this issue, we propose a UAV drone swarm cluster-based data collection (USCDC) algorithm. Our algorithm adopts a two-layer UAV drone architecture: cooperative UAV drones are deployed at the bottom layer to collect data from sensor clusters within each sub-region, and a sink UAV drone is configured at the top layer to aggregate data from all cooperative UAV drones and transmit it to the base station. This hierarchical cooperation mechanism allows cooperative UAV drones to offload data without returning to the base station, significantly reducing path overhead and energy consumption. In this paper, we detail the system model, problem formulation, and proposed algorithms, including a temporal cooperative scheduling scheme, an improved ant colony optimization for regional path planning, and an improved particle swarm optimization for sink UAV trajectory optimization. Extensive simulations demonstrate that our USCDC algorithm reduces total path length by 52% and energy consumption by 35% compared to the baseline parallel data collection approach.
Introduction
The rapid proliferation of Internet of Things (IoT) applications, such as environmental monitoring, smart agriculture, and industrial automation, has led to the deployment of massive wireless sensor networks. In many scenarios, sensors are spatially distributed across large areas and may be separated by natural or artificial barriers, forming isolated clusters. Traditional data collection methods relying on multi-hop relaying become inefficient or infeasible due to energy constraints and connectivity issues. UAV drones, with their high mobility and on-demand deployment, have emerged as a promising solution for data collection in such fragmented environments. However, when multiple UAV drones are deployed, the lack of effective coordination results in each UAV drone independently flying back to the base station after collecting data, causing redundant paths and wasted energy. Our work focuses on minimizing the total energy consumption of the UAV drone swarm by introducing a cluster-based hierarchical architecture.
System Model
We consider a large-scale WSN consisting of several isolated sub-regions. Each sub-region contains a set of sensor nodes that form clusters using a distributed clustering algorithm (e.g., LEACH). In each round, cluster heads are elected to aggregate data from their members. The set of cluster heads in the o-th sub-region is denoted as Zo = {zo,1, zo,2, …, zo,A}, and the data volume collected by cluster head zo,a is Do,a. All sensor nodes have the same initial energy En and communication radius Rn.
The UAV drone swarm consists of two layers. The bottom layer comprises W cooperative UAV drones U = {u1, u2, …, uW}, each assigned to one sub-region. Each cooperative UAV drone flies at height Hc with speed Vc and has initial energy Ec. The top layer has one sink UAV drone us that collects data from all cooperative UAV drones and finally returns to the base station. The sink UAV drone flies at height Hs with speed Vs and has initial energy Es.
The communication between a cooperative UAV drone and a cluster head uses a line-of-sight (LoS) model. The path loss is given by:
$$
l_{\text{LoS}}^{u_j, z_{o,a}} = 20 \lg\left(\frac{4\pi f_c d_{u_j, z_{o,a}}}{c}\right) + \eta_{\text{LoS}}
$$
where fc is the carrier frequency, c is the speed of light, d is the distance, and ηLoS is the excess loss due to LoS. The channel gain is:
$$
h_{u_j, z_{o,a}} = 10^{-l_{\text{LoS}}^{u_j, z_{o,a}}/10}
$$
The data rate is computed using Shannon’s formula:
$$
r_{u_j, z_{o,a}} = B \log_2\left(1 + \frac{P_t h_{u_j, z_{o,a}}}{N_0 B}\right)
$$
Similarly, the air-to-air communication between the sink UAV drone and a cooperative UAV drone also follows the LoS model.
The energy consumption of a UAV drone consists of three parts: hovering power, horizontal flight power, and communication power. The hovering power Ph is:
$$
P_h = \sqrt{\frac{(m_{\text{tot}} g)^3}{2\pi \sigma_p^2 \delta_p \rho}}
$$
where mtot is the total mass, g is gravity, σp is propeller radius, δp is number of propellers, and ρ is air density. The flight power Pf is linear with flight speed vf:
$$
P_f = \frac{P_{\max} – P_{\text{idle}}}{v_{\max}} v_f + P_{\text{idle}}
$$
For a cooperative UAV drone uj in round x, the total energy is:
$$
E_{u_j}^x = E_{u_j,\text{fly}}^x + E_{u_j,\text{coll}}^x + E_{u_j,\text{wait}}^x + E_{u_j,\text{trans}}^x
$$
where the flight energy is:
$$
E_{u_j,\text{fly}}^x = \frac{L_{u_j}^x}{V_f} (P_f + P_h)
$$
the data collection energy is:
$$
E_{u_j,\text{coll}}^x = \frac{D_{u_j}^x}{r_{u_j,z_{o,a}}} (P_h + P_{\text{com}})
$$
the waiting energy is:
$$
E_{u_j,\text{wait}}^x = t_{\text{wait}}^x P_h
$$
and the transmission energy to sink is:
$$
E_{u_j,\text{trans}}^x = \frac{D_{u_j}^x}{r_{u_s,u_j}} (P_h + P_{\text{com}})
$$
For the sink UAV drone us in round x:
$$
E_{u_s}^x = E_{u_s,\text{fly}}^x + E_{u_s,\text{trans}}^x
$$
with flight energy:
$$
E_{u_s,\text{fly}}^x = \frac{S_{u_s}^x}{V_s} P_f
$$
and transmission energy:
$$
E_{u_s,\text{trans}}^x = \frac{D_{u_s}}{r_{u_s,u_j}} (P_h + P_{\text{com}})
$$
The total energy consumption over xmax rounds is:
$$
E_{\text{total}} = \sum_{x=1}^{x_{\max}} \left( \sum_{j=1}^{W} E_{u_j}^x + E_{u_s}^x \right)
$$
Problem Formulation
Our objective is to minimize the total energy consumption subject to constraints on UAV drone energy budgets, regional boundaries, round-trip requirements for the sink UAV drone, and complete data collection. This is a non-convex optimization problem. We decompose it into two subproblems: (1) path planning for cooperative UAV drones within each sub-region, and (2) trajectory optimization for the sink UAV drone visiting the cooperative UAV drones. We solve these subproblems using heuristic algorithms.
Proposed Algorithms
UAV Drone Temporal Cooperative Scheduling (UTCS)
To coordinate the timing of cooperative and sink UAV drones, we propose UTCS. The sink UAV drone uses absolute time while cooperative UAV drones use relative time. The algorithm determines the departure time of the sink UAV drone as:
$$
t_d = \max_{j \in W} t_{u_j}^{\text{coll}}
$$
where tu_jcoll is the time when cooperative UAV drone uj finishes data collection. The waiting times for each cooperative UAV drone are:
$$
t_{u_j}^{\text{wait1}} = \max(0, t_{u_j}^{\text{us,arr}} – t_{u_j}^{\text{coll}})
$$
$$
t_{u_j}^{\text{wait2}} = \max(0, t_{\text{sys,end}} – t_{u_j}^{\text{trans,end}})
$$
This mechanism synchronizes the swarm and minimizes idle energy.
Improved Ant Colony Optimization (IACO) for Cooperative UAV Drone Path
We enhance the classic ant colony algorithm by incorporating an elite ant strategy and a position state transfer mechanism. The transition probability from cluster head a to b is:
$$
p_{ab}^k(t) = \frac{[\tau_{ab}(t)]^\alpha [\varepsilon_{ab}(t)]^\beta}{\sum_{i \in \text{allowed}_k} [\tau_{ab}(t)]^\alpha [\varepsilon_{ab}(t)]^\beta}
$$
Pheromone update uses elite ants: among the top λ proportion of ants, extra pheromone is deposited. The update rule is:
$$
\Delta \tau^k_{ab} =
\begin{cases}
(1+\xi)Q/C_k, & \text{if } k \in \lambda, (a,b) \in F_k \\
Q/C_k, & \text{if } (a,b) \in F_k \\
0, & \text{otherwise}
\end{cases}
$$
The position state transfer mechanism uses the end position of the previous round as the start position of the next round:
$$
T_{\text{start}}^x = T_{\text{end}}^{x-1} = (X_{\text{end}}^{x-1}, Y_{\text{end}}^{x-1}, H_c)
$$
This reduces inter-round wasted flight.
Improved Particle Swarm Optimization (IPSO) for Sink UAV Drone Trajectory
We adapt the standard PSO with adaptive inertia weight and a bi-objective fitness function. The inertia weight is:
$$
w = w_{\max} – (w_{\max} – w_{\min}) \cdot \frac{\kappa_{\text{iter}}}{M}
$$
The fitness function for a particle representing a visiting sequence is:
$$
\psi = w_1 \frac{d}{d_{\max}} + w_2 \frac{e}{e_{\max}}
$$
where d and e are path length and energy, and dmax, emax are the maximum values in the population. This balances both objectives.
Simulation Results
We implemented our algorithms in PyCharm and compared USCDC with a baseline parallel data collection (USPDC) where each cooperative UAV drone returns to base after collection (no sink UAV drone). The simulation area is 2000 m × 2000 m with 6 sub-regions, each containing 100 sensor nodes. Key parameters are listed in the following table.
| Parameter | Value |
|---|---|
| Number of sub-regions (W) | 6 |
| Sub-region radius (Rr) | 200 m |
| Node initial energy (En) | 1 J |
| Node communication radius (Rn) | 50 m |
| Sink UAV initial energy (Es) | 2,700,000 J |
| Sink UAV height (Hs) | 75 m |
| Sink UAV speed (Vs) | 25 m/s |
| Cooperative UAV initial energy (Ec) | 1,500,000 J |
| Cooperative UAV height (Hc) | 50 m |
| Cooperative UAV speed (Vc) | 15 m/s |
| UAV flight power (Pf) | 90 W |
| UAV hovering power (Ph) | 30 W |
| UAV communication power (Pcom) | 0.5 W |
| Number of ants (nant) | 100 |
| α (pheromone importance) | 1.5 |
| β (heuristic importance) | 3.0 |
| µ (pheromone decay) | 0.6 |
| Q (pheromone intensity) | 100 |
| Number of particles (npart) | 100 |
| wmax | 0.95 |
| wmin | 0.4 |
| θ1 (individual learning factor) | 2.05 |
| θ2 (social learning factor) | 2.05 |
We conducted 100 independent runs and averaged the results. The following figure shows the total path length per round for USCDC and USPDC.
From the results, USCDC achieved a total path length of approximately 10 km per round, while USPDC required about 21 km—a reduction of 52.4%. This is because cooperative UAV drones in USCDC only fly within their local sub-regions and the sink UAV drone efficiently aggregates data without each UAV drone returning to the base station.
The energy consumption per round is shown in the following analysis. USCDC consumed about 8.5×104 J per round, whereas USPDC consumed 13.5×104 J, a reduction of 37%. The energy efficiency (bits per joule) was also improved: USCDC achieved 10 bit/J compared to 6.5 bit/J for USPDC, a 53.8% improvement.
We also evaluated scalability under different WSN sizes: small (1000 m × 1000 m, 3 sub-regions), medium (2000 m × 2000 m, 6 sub-regions), and large (3000 m × 3000 m, 12 sub-regions). The total path length comparison is summarized in the next table.
| WSN Size | USPDC | USCDC | Reduction |
|---|---|---|---|
| Small | 12.5 | 9.0 | 28% |
| Medium | 21.0 | 10.0 | 52% |
| Large | 38.5 | 13.0 | 66.3% |
Energy consumption under different sizes is shown below:
| WSN Size | USPDC | USCDC | Reduction |
|---|---|---|---|
| Small | 2.1 | 1.66 | 21% |
| Medium | 4.05 | 2.63 | 35% |
| Large | 7.8 | 4.68 | 40% |
The results confirm that our USCDC algorithm scales well; as the network grows, the relative advantage over USPDC increases. The energy saving is slightly less than path saving due to increased waiting times for cooperative UAV drones, which adds hovering energy. This suggests that further optimization of the sink UAV drone’s visit order could reduce waiting times.
Conclusion
We have presented a UAV drone swarm cluster-based data collection algorithm for wireless sensor networks. By employing a two-layer architecture with cooperative UAV drones and a sink UAV drone, our USCDC algorithm significantly reduces path length and energy consumption compared to traditional parallel collection without clustering. The temporal cooperative scheduling, improved ant colony optimization, and improved particle swarm optimization work together to enhance efficiency. Simulation results show up to 52% reduction in path length and 35% reduction in energy consumption, with excellent scalability. Future work will focus on optimizing the sink UAV drone’s visiting sequence to minimize waiting times and on extending the algorithm to continuous sensor distributions.