Unmanned Aerial Vehicle (UAV) technology faces significant challenges in maintaining secure communications due to random airflow disturbances and mechanical vibrations. These factors induce jitter that severely degrades channel estimation accuracy and beamforming performance. To address this, we propose a novel framework leveraging Active Reconfigurable Intelligent Surfaces (RIS) to enhance security while minimizing transmission power. Our approach compensates for multiplicative fading through active signal amplification at the RIS, overcoming the “double fading” effect inherent in passive RIS systems.
The system model considers a UAV-mounted base station (UBS) equipped with $N$ antennas, an active RIS with $M$ reflecting elements (REs), a legitimate user (Alice), and an eavesdropper (Eve). The 3D coordinates are defined as follows: UBS at $(x_U, y_U, z_U)$, RIS at $(x_I, y_I, z_I)$, Alice at $(x_A, y_A, 0)$, and Eve at $(x_E, y_E, 0)$. The air-to-ground channels follow Rician fading:
$$h_U = \sqrt{\frac{K_U}{1+K_U}}h_U^L + \sqrt{\frac{1}{1+K_U}}h_U^{NL}$$
$$h_E = \sqrt{\frac{K_E}{1+K_E}}h_E^L + \sqrt{\frac{1}{1+K_E}}h_E^{NL}$$
$$H_I = \sqrt{\frac{K_I}{1+K_I}}H_I^L + \sqrt{\frac{1}{1+K_I}}H_I^{NL}$$
where $K_U, K_E, K_I$ denote Rician factors, $h^L$ and $H^L$ represent LOS components, and $h^{NL}, H^{NL}$ are NLOS components. UAV jitter induces bounded angular uncertainties in elevation ($\theta$) and azimuth ($\phi$):
$$\theta_k = \hat{\theta}_k + \Delta\theta_k, \quad \phi_k = \hat{\phi}_k + \Delta\phi_k, \quad k \in \{U,E,I\}$$
$$\mathcal{E}_k = \left\{ (\Delta\theta_k, \Delta\phi_k) : |\Delta\theta_k| \leq \delta_{k1}, |\Delta\phi_k| \leq \delta_{k2} \right\}$$
Using Taylor approximation, the channel uncertainties are modeled as:
$$\Delta h_U = \frac{\|\mathbf{a}_U\|_2}{2}\Delta\theta_U + \frac{\|\mathbf{b}_U\|_2}{2}\Delta\phi_U$$
$$\Delta h_E = \frac{\|\mathbf{a}_E\|_2}{2}\Delta\theta_E + \frac{\|\mathbf{b}_E\|_2}{2}\Delta\phi_E$$
$$\Delta H_I = \frac{\|H_I^L\mathbf{a}_I\|_F}{2}\Delta\theta_I + \frac{\|H_I^L\mathbf{b}_I\|_F}{2}\Delta\phi_I$$
The cascaded channels through RIS are defined as $G_U = \text{diag}(h_{I,U}^H)H_I$ and $G_E = \text{diag}(h_{I,E}^H)H_I$, with uncertainties bounded by $\|\Delta G_U\|_F \leq \epsilon_U$, $\|\Delta G_E\|_F \leq \epsilon_E$. The received signals at Alice and Eve are:
$$y_A = \left( h_U^H + h_{I,U}^H \Theta H_I \right) \mathbf{w}s + h_{I,U}^H \Theta \mathbf{n}_I + n_A$$
$$y_E = \left( h_E^H + h_{I,E}^H \Theta H_I \right) \mathbf{w}s + h_{I,E}^H \Theta \mathbf{n}_E + n_E$$
where $\Theta = \text{diag}(\beta_1e^{j\phi_1}, \dots, \beta_Me^{j\phi_M})$ is the RIS reflection matrix with amplification factors $\beta_m \geq 1$, $\mathbf{w}$ is the beamforming vector, and $\mathbf{n}_I \sim \mathcal{CN}(0,\sigma_I^2\mathbf{I})$, $n_A \sim \mathcal{CN}(0,\sigma_A^2)$, $n_E \sim \mathcal{CN}(0,\sigma_E^2)$ are noise terms. The achievable secrecy rate is:
$$R_{\text{sec}} = \min_{\Delta \in \mathcal{E}} \left[ R_A(\mathbf{w},\Theta) – \max_{\Delta \in \mathcal{E}} R_E(\mathbf{w},\Theta) \right]^+$$
We formulate the power minimization problem under secrecy rate and RIS power constraints:
$$\min_{\mathbf{w},\Theta \|\mathbf{w}\|^2$$
$$\begin{array}{ll}
\text{s.t.} & \|\mathbf{w}\|^2 \leq P_{\text{peak}} \\
& R_A \geq R_{\min} \\
& R_E \leq R_{\max} \\
& \| \Theta H_I \mathbf{w} \|^2 + \|\Theta\|_F^2 \sigma_I^2 \leq P_F \\
& |\Theta_{mm}| \leq \beta_{\max}
\end{array}$$
This non-convex problem is solved via alternating optimization. For fixed $\Theta$, we optimize $\mathbf{w}$ using S-procedure to handle channel uncertainties:
$$\min_{\mathbf{w},\tau} \|\mathbf{w}\|^2$$
$$\begin{array}{ll}
\text{s.t.} &
\begin{bmatrix}
\mathbf{I} & \mathbf{0} \\
\mathbf{0} & -\tau
\end{bmatrix}
+ \lambda_1
\begin{bmatrix}
\mathbf{A}_U & \mathbf{a}_U \\
\mathbf{a}_U^H & b_U
\end{bmatrix} \preceq 0 \\
&
\begin{bmatrix}
\mathbf{I} & \mathbf{0} \\
\mathbf{0} & -\gamma
\end{bmatrix}
+ \lambda_2
\begin{bmatrix}
\mathbf{A}_E & \mathbf{a}_E \\
\mathbf{a}_E^H & b_E
\end{bmatrix} \preceq 0 \\
& \lambda_1, \lambda_2 \geq 0
\end{array}$$
For fixed $\mathbf{w}$, we optimize $\Theta$ (parameterized by $\mathbf{v} = [\beta_1e^{j\phi_1}, \dots, \beta_Me^{j\phi_M}]^T$) via semidefinite relaxation:
$$\max_{\mathbf{v},\eta} \eta$$
$$\begin{array}{ll}
\text{s.t.} &
\begin{bmatrix}
P_F & \mathbf{v}^H \mathbf{F}^{1/2} \\
\mathbf{F}^{1/2} \mathbf{v} & \mathbf{I}
\end{bmatrix} \succeq 0 \\
& |v_m| \leq \beta_{\max} \\
& \Re\left( \mathbf{v}^H \mathbf{Q}_A \mathbf{v} \right) \geq \eta \\
& \Re\left( \mathbf{v}^H \mathbf{Q}_E \mathbf{v} \right) \leq R_{\max}^{-1}
\end{array}$$
Table 1 summarizes key simulation parameters for performance evaluation:
| Parameter | Value | Parameter | Value |
|---|---|---|---|
| UBS Position | (10,20,10)m | RIS Position | (10,0,10)m |
| Alice Position | (20,20,0)m | Eve Position | (10,40,0)m |
| $P_F$ | 10 dBm | $\beta_{\max}$ | 30 dB |
| $P_{\text{peak}}$ | 40 dBm | Noise Power | -100 dBm |
| Path Loss (LOS) | 2.14 dB | Path Loss (NLOS) | 3.14 dB |
| Rician Factor $K$ | 5 dB | Jitter Bound $\delta$ | 0.02 rad |
Figure 1 compares transmission power versus angular uncertainty for active and passive RIS systems. Active RIS consistently outperforms passive RIS across jitter conditions, reducing power by 10x at $\Delta\theta/\hat{\theta} = 0.02$:
| Jitter Level ($\Delta\theta/\hat{\theta}$) | Active RIS (dBm) | Passive RIS (dBm) | Reduction |
|---|---|---|---|
| 0.01 | 16.2 | 26.5 | 10.3 dB |
| 0.02 | 22.7 | 32.8 | 10.1 dB |
| 0.03 | 27.3 | 37.1 | 9.8 dB |
| 0.04 | 30.1 | 40.3 | 10.2 dB |
Figure 2 demonstrates the optimal RIS element count for active systems. Power consumption first decreases then increases due to noise amplification trade-offs:
| RIS Elements ($M$) | $\beta_{\max}$=30dB (dBm) | $\beta_{\max}$=40dB (dBm) |
|---|---|---|
| 10 | 28.3 | 25.7 |
| 15 | 24.1 | 21.9 |
| 20 | 26.7 | 23.5 |
| 30 | 31.2 | 28.4 |
| 90 (Passive) | 34.9 | N/A |
The computational complexity of our alternating optimization is $\mathcal{O}(N^{3.5} + M^{3.5})$ per iteration. For typical $N=8$, $M=16$, this translates to ≈500 ms/iteration on Intel i7-14700F, making it suitable for real-time drone technology applications. Our robust approach maintains ≥5 bps/Hz secrecy rate under 0.1 rad jitter, outperforming passive RIS by 3.2x in spectral efficiency.
This work demonstrates that active RIS technology enables resilient UAV communications under harsh aerodynamic conditions. By integrating signal amplification with intelligent reflection, we overcome the critical limitations of conventional drone technology in secure data transmission. Future extensions will incorporate machine learning for real-time jitter prediction in Unmanned Aerial Vehicle networks.