Enhancing reliable and energy-efficient UAV communications with RIS and deep reinforcement learning

EMI and signal quality in UAV-RIS systems

Improving signal quality through UAV-based optimizations traditionally focuses on path planning, coverage extension, and signal amplification. In particular, DRL-based UAV optimization strategies are designed to enhance line-of-sight (LoS) communication, optimize trajectory planning, and adjust altitudes to maximize signal strength and minimize path loss (Du & Ma, 2024; Mahalle et al., 2024). These optimizations are effective in increasing SINR by optimizing signal propagation paths and reducing multi-path fading. However, enhancing signal quality alone does not inherently address the challenges posed by EMI, which can significantly degrade communication reliability. EMI, particularly from high-power components like GaN power amplifiers, introduces noise and signal distortions that disrupt phase alignment and amplitude stability, directly affecting SINR regardless of path optimization or trajectory planning (Du & Ma, 2024).

While conventional DRL-based UAV optimizations focus on enhancing LoS and boosting signal strength, they are generally not equipped to counteract the impact of EMI. Electromagnetic interference disrupts signal transmission through unanticipated noise spikes and phase distortions, leading to packet losses and increased bit error rates (BER) even in optimized path scenarios. Mahalle et al. (2024) demonstrated that shielding techniques and adaptive filtering could reduce EMI-induced distortions, but these methods are primarily static and lack real-time adaptability. Our approach introduces a novel DRL-based adaptation mechanism that actively senses EMI fluctuations and optimally adjusts both UAV positioning and RIS phase shifts to mitigate interference. Unlike traditional optimization methods, this framework continuously learns and adapts to changing interference patterns, ensuring both enhanced signal quality and effective EMI mitigation in real-time. Table 1 shows the comparative analysis of previous studies.

The existing research on RIS-assisted UAV communications using deep reinforcement learning reveals notable advancements in optimizing resource allocation, energy efficiency, and security across various applications. However, significant gaps remain in addressing the real-time adaptability of these systems under dynamic environmental conditions and multi-user interference. Current studies primarily focus on theoretical and simulation-based models, often overlooking practical deployment challenges such as computational complexity and scalability in large-scale networks. Furthermore, there is limited exploration into integrating advanced machine learning techniques for more robust decision-making capabilities, particularly in heterogeneous network environments. Addressing these gaps is essential for advancing the practical implementation and effectiveness of RIS-assisted UAV communications.

System model: UAV-RIS-assisted communication

The system under consideration consists of multiple UAVs deployed to assist communication between a base station (BS) and multiple ground users. RIS are installed to enhance signal quality by reflecting and steering signals toward desired users, thus improving signal strength, coverage, and overall communication quality.

Influence of GaN power amplifier EMI on UAV-RIS system

GaN power amplifiers, commonly used in wireless communication systems due to their high efficiency and power output, generate significant EMI that can degrade signal quality and disrupt communication. The complete flow is shows in Fig. 2.

EMI modeling

The electromagnetic interference introduced by GaN power amplifiers affects both direct and reflected communication links. We model the EMI power as ${\mathrm{P}}_{\mathrm{E}\mathrm{M}\mathrm{I}}$, which contributes to the overall interference in the system. The received EMI power at the $\mathrm{j}$-th user is given by:

(3) $${\mathrm{P}}_{\mathrm{E}\mathrm{M}\mathrm{I},\mathrm{j}}=\sum _{\mathrm{i}=1}^{\mathrm{U}}{\mathrm{\beta }}_{\mathrm{i},\mathrm{j}}{\mathrm{P}}_{\mathrm{G}\mathrm{a}\mathrm{N},\mathrm{i}}$$where:

• ${\mathrm{P}}_{\mathrm{G}\mathrm{a}\mathrm{N},\mathrm{i}}$ is the EMI generated by the $\mathrm{i}$-th UAV’s GaN power amplifier.

• ${\mathrm{\beta }}_{\mathrm{i},\mathrm{j}}$ is the interference factor that accounts for the propagation loss of EMI from the $\mathrm{i}$-th UAV to the $\mathrm{j}$-th user.

This additional interference negatively impacts the SINR for each user. The modified SINR considering GaN EMI becomes:

(4) $${\mathrm{S}\mathrm{I}\mathrm{N}\mathrm{R}}_{\mathrm{i},\mathrm{j}}^{\mathrm{E}\mathrm{M}\mathrm{I}}={\displaystyle \frac{{\mathrm{P}}_{{\mathrm{u}}_{\mathrm{i}}}{\left|{\mathrm{h}}_{\mathrm{i},\mathrm{j}}\right|}^2}{{\mathrm{N}}_0+{\sum }_{\mathrm{m}\ne \mathrm{i}}{\mathrm{P}}_{{\mathrm{u}}_{\mathrm{m}}}{\left|{\mathrm{h}}_{\mathrm{m},\mathrm{j}}\right|}^2+{\mathrm{I}}_{\mathrm{i},\mathrm{j}}+{\mathrm{P}}_{\mathrm{E}\mathrm{M}\mathrm{I},\mathrm{j}}}}$$QPSK is an efficient modulation technique widely used in wireless communication systems to increase spectral efficiency and improve signal robustness. When integrated into UAV-RIS communication systems, QPSK enhances the system’s ability to transmit more data using the same bandwidth, which is critical in mitigating the effects of EMI generated by GaN power amplifiers.

In this work, we integrate QPSK modulation into the UAV-RIS system to further combat the adverse effects of EMI. QPSK offers the following advantages in this context:

• Improved spectral efficiency: QPSK allows for the transmission of two bits per symbol, effectively doubling the data rate compared to binary phase shift keying (BPSK). This results in higher throughput and improved communication performance, particularly in environments where EMI is prevalent.

• Resilience to EMI: By modulating the phase of the carrier signal, QPSK can effectively differentiate between signal components even when interference from GaN power amplifiers is present. This helps in maintaining a reliable communication link in challenging environments.

• Signal integrity: The modulation technique is less susceptible to signal degradation, allowing for clearer and more reliable transmission even in the presence of high-power EMI.

Incorporating QPSK into the system modifies the SINR calculation, as the system now benefits from more robust signal transmission. The modified SINR expression, taking into account both GaN EMI and QPSK modulation, is given by:

(5) $${\mathrm{S}\mathrm{I}\mathrm{N}\mathrm{R}}_{\mathrm{i},\mathrm{j}}^{\mathrm{Q}\mathrm{P}\mathrm{S}\mathrm{K}}={\displaystyle \frac{\left({\mathrm{P}}_{\mathrm{u}\mathrm{i}}|{\mathrm{h}}_{\mathrm{i},\mathrm{j}}|2\right)}{\left(\mathrm{N}0+{\sum }_{\mathrm{n}=0}^{\mathrm{m}}\mathrm{M}=\mathrm{i}{\mathrm{P}}_{\mathrm{u}\mathrm{m}}|{\mathrm{h}}_{\mathrm{m},\mathrm{j}}|2+{\mathrm{I}}_{\mathrm{i},\mathrm{j}}+\mathrm{P}\mathrm{E}{\mathrm{M}}_{\mathrm{I},\mathrm{j}}\right)}}$$where:

• PEM_{i, j} represents the EMI power received by the j-th user due to the GaN power amplifier.

• I_{i, j} is the interference from other users, and h_{i, j} is the channel gain between the UAV-RIS and the j-th user.

By integrating QPSK into the UAV-RIS system, the SINR can be enhanced, improving signal quality and reducing the impact of EMI. This integration is particularly useful in scenarios where the UAVs and RIS face severe interference from GaN power amplifiers, enabling the system to maintain high data transmission rates and reliable communication links. The combination of QPSK with UAV-RIS-DRL provides a robust solution for optimizing spectral efficiency and mitigating the negative effects of EMI. Figure 3 shows the QPSK-Enhanced SINR Calculation.

The challenge here is to mitigate the impact of ${\mathrm{P}}_{\mathrm{E}\mathrm{M}\mathrm{I},\mathrm{j}}$ using intelligent UAV placement and RIS phase shift configurations.

Proposed DRL-based optimization model with QPSK

To dynamically optimize UAV positions and RIS configurations in the presence of EMI, we propose a DRL approach. The DRL agent learns from the environment, continuously adjusting UAV placements and RIS phase shifts to minimize the effect of EMI while maximizing SINR and energy efficiency. Figure 4 below shows the proposed model flow.

System performance evaluation

To evaluate the proposed framework, we simulate various communication scenarios with different levels of EMI. The system performance is measured in terms of:

• SINR: Improvement in SINR with and without the influence of GaN EMI.

• Energy efficiency: The trade-off between energy consumption and communication quality.

Energy efficiency in the proposed DRL-based UAV-RIS framework is defined as the ratio of effective data transmission (in bits) to the total energy consumption (in Joules) across UAV and RIS operations. Mathematically, it can be expressed as:

(11) $$\eta ={\displaystyle \frac{D_{transmitted}}{E_{total}}}$$where $D_{transmitted}$ represents the total data successfully transmitted, and $E_{total}$ is the cumulative energy consumption of UAV propulsion, signal amplification, RIS phase shifting, and DRL computation overhead. This formulation allows for assessing the communication efficiency in terms of data throughput relative to energy expenditure, making it particularly relevant in EMI-prone environments where signal correction and adaptive adjustments are required.

• EMI impact mitigation: Reduction in the adverse effects of EMI on system performance.

Signal-to-interference-plus-noise ratio

The SINR performance of the proposed system was evaluated under different communication scenarios, with and without the influence of EMI. The DRL-optimized UAV-RIS system demonstrated a significant improvement in SINR compared to the baseline system, particularly in environments affected by EMI. Table 2 presents the SINR values for both systems across various user positions.

From the results in Table 2, it is evident that the proposed UAV-RIS system consistently outperforms the baseline system, even in the presence of EMI. The degradation in SINR due to EMI is mitigated by the intelligent placement of UAVs and the dynamic adjustment of RIS phase shifts using DRL. The average SINR improvement is approximately 6.5 dB without EMI and 4.0 dB with EMI. The Comparison of baseline and proposed UAV-RIS system in term of SINR under EMI is shown in Fig. 5.

Energy efficiency

The energy efficiency of the proposed system is compared with the baseline system in Table 3. The energy efficiency of the UAV-RIS system is significantly higher, demonstrating the ability to maintain robust communication while consuming less power.

The proposed UAV-RIS system achieves up to 38% higher energy efficiency compared to the baseline, even in the presence of EMI. This is primarily due to the optimized UAV positioning and the enhanced signal steering capabilities of the RIS, which reduces the energy required for communication over longer distances and in high-interference environments. The energy efficiency comparison is shown in Fig. 6.

EMI impact mitigation

The proposed system’s ability to mitigate the adverse effects of EMI is one of its key strengths. Table 4 provides a quantitative analysis of EMI impact mitigation, showing the percentage reduction in SINR degradation when EMI is present.

The results in Table 4 show that the proposed system reduces the EMI impact by over 70% compared to the baseline system, highlighting the effectiveness of DRL in mitigating interference. Moreover, the Fig. 7 shows the EMI impact mitigation in baseline and proposed system.

The proposed DRL-optimized UAV-RIS framework demonstrates a significant capability to mitigate EMI impacts, especially those induced by high-power GaN power amplifiers. As illustrated in Table 4 and Fig. 7, the system achieves a 70% reduction in SINR degradation when exposed to interference from GaN power amplifiers. This notable improvement is attributed to the real-time adaptability of UAV positions and RIS phase shifts, orchestrated by the DRL agent. By dynamically optimizing signal paths, the system effectively minimizes electromagnetic disruptions, resulting in enhanced communication reliability and stability.

Comparatively, traditional UAV-RIS systems lack dynamic interference management capabilities, relying predominantly on static configurations that are ineffective against fluctuating EMI conditions. Prior studies have focused on shielding and filtering techniques to reduce EMI, but these methods are often limited by their inability to respond to real-time changes in signal interference (Du & Ma, 2024; Mahalle et al., 2024). In contrast, the proposed DRL-based approach enables continuous learning and adaptation, allowing UAVs and RIS configurations to adjust to interference patterns in real-time. This advancement is particularly impactful in high-density urban deployments, where EMI from various electronic sources is prevalent, often leading to severe communication disruptions. The experimental results indicate that even under intense interference scenarios, the proposed framework maintains a consistent SINR improvement of 6.5 dB, while extending coverage area by 35% and increasing energy efficiency by 38%.

Furthermore, the integration of QPSK modulation enhances spectral efficiency, allowing for robust data transmission despite EMI influences. This modulation technique, combined with real-time DRL optimization, ensures that communication quality is preserved, even as environmental interference fluctuates. These results highlight the effectiveness of the proposed UAV-RIS framework not only in maintaining signal quality but also in extending communication reliability to regions with high electromagnetic pollution.

Latency

Latency is a critical metric, particularly for real-time communication applications. Table 5 compares the latency of the baseline system and the proposed UAV-RIS system across various user positions. The latency comparison is also shows in Fig. 8.

From Table 5, it can be observed that the proposed UAV-RIS system reduces latency by an average of 24%, even in the presence of EMI. This is achieved by optimizing the signal paths and minimizing interference.

Latency in this study is defined as the total time taken for data packets to travel from the base station to the end-users through the UAV-RIS-assisted communication link. It includes the propagation delay, processing delay at the RIS, and UAV signal relaying delay. The expression for latency $\left(L\right)$ is given by:

(12) $$L=T_{propagation}+T_{processing}+T_{relay}$$where:

• $T_{propagation}$ is the time for the signal to travel through the medium,

• $T_{processing}$ is the delay introduced by phase adjustments at RIS, and

• $T_{relay}$ is the time taken for UAV to process and forward the signal.

The DRL framework optimizes these paths to minimize latency, ensuring efficient communication even in high-interference conditions.

Coverage area

The coverage area of the UAV-RIS system is larger compared to the baseline, as shown in Table 6. This increase in coverage is particularly useful in rural and remote areas where traditional communication infrastructure is limited.

The results from the Fig. 9 indicate that the UAV-RIS system can extend coverage by up to 35%, providing more extensive communication reach, especially in environments where direct line-of-sight is obstructed.

In addition to the key performance improvements observed in SINR, energy efficiency, and EMI mitigation, we further analyzed the system’s performance by evaluating the output power levels (in dBm), error counts, and error rates. Table 7 illustrates the error count and error rate as the output power of the system increases. The results from Fig. 10 show that for lower output power levels (below 28 dBm), the system maintains zero error counts with no observable error rates, indicating highly reliable communication. However, as the output power exceeds 31 dBm, the system begins to experience increased error counts, with the error rate rising progressively from 2.44 × 1 0 − 4 2.44 × 10 −4 to 7.36 × 1 0 − 3 7.36 × 10 −3 as the output power reaches 33.47 dBm. This behavior can be attributed to the influence of noise and interference, particularly as the system operates at higher power levels, where nonlinearities and EMI from GaN power amplifiers become more pronounced. The DRL-optimized framework demonstrated an ability to mitigate these effects by dynamically adjusting the UAV positions and RIS configurations, but the observed increase in errors highlights the inherent limitations of the system when operating at very high power levels. Despite this, the QPSK modulation scheme provided a notable level of resilience, with the system maintaining a relatively low error rate of up to 33 dBm. This suggests that QPSK, in conjunction with the UAV-RIS-DRL framework, is effective at maintaining communication integrity in most operational power ranges. Further refinement of the DRL algorithm and additional interference mitigation techniques could potentially reduce the error rates at higher power levels, making the system even more robust.

Table 7 presents the error count and error rate alongside the corresponding output power levels.

Comparison with previous studies

To contextualize the performance of our proposed DRL-based UAV-RIS integration, we compare the results with those from previous studies. The comparison is summarized in Table 8.

The comparison in Table 8 shows that our proposed DRL-based UAV-RIS integration outperforms previous studies across all key metrics. Specifically, our study demonstrates a 20.0% improvement in SINR in urban scenarios, a 16.7% improvement in energy efficiency in suburban environments, and a 9.2% reduction in latency in rural areas. These results validate the effectiveness of our proposed system and underscore the advantages of integrating DRL for dynamic optimization in complex communication networks. The results presented in this section highlight the substantial improvements in system performance achieved through the integration of DRL with UAV-RIS and QPSK simulation. The DRL-based model consistently outperforms non-DRL approaches across all key metrics, including SINR, energy efficiency, latency, and coverage area. Furthermore, our proposed system demonstrates significant gains compared to previous studies, confirming its potential to enhance communication quality and efficiency in diverse environments.

Validation of results with theoretical analysis

To validate the simulation results presented in the preceding Results and Discussions sections about the key metrics, comparisons were conducted with theoretical models for both SINR and latency. These theoretical values were derived from standard path loss models, Rician fading assumptions, and EMI interference calculations, providing a benchmark for evaluating the accuracy of the proposed DRL-based UAV-RIS optimization framework.

Theoretical SINR calculation

The theoretical SINR for the UAV-RIS communication system was computed using the standard path loss model combined with Rician fading effects, represented as follows:

(13) $$SINR-{\displaystyle \frac{P_s.|\mathrm{h}{|}^2}{I+N}}$$where:

• $P_s$ is the transmitted signal power,

• $|\mathrm{h}{|}^2$ is the channel gain modeled with Rician fading,

• I is the interference power, including EMI from GaN amplifiers, and

• $N$ is the noise power.

For urban environments, the Rician K-factor was set to 8, reflecting stronger line-of-sight (LoS) components. Suburban environments assumed a K-factor of 5, and rural environments used 3 to account for lower LoS dominance. These configurations ensured the theoretical SINR calculations were consistent with realistic signal propagation conditions.

Table 9 presents the comparison of the simulated SINR and latency values against the theoretical calculations. The results demonstrate a 5–7% deviation, indicating high reliability and robustness of the DRL-based optimization framework in managing interference, maintaining communication quality, and reducing latency. The minimal deviation highlights the system’s accuracy in dynamically adjusting UAV positioning and RIS phase shifts to counteract the effects of EMI, validating the simulation outcomes with strong theoretical alignment.

The close alignment of theoretical and simulated results validates the effectiveness of the DRL-based UAV-RIS framework, particularly in optimizing SINR and reducing latency in real-time. This theoretical comparison supports the robustness of the proposed system in managing EMI while maintaining energy-efficient, high-quality communication links.

Influence of RIS reflective elements on system performance

The number of reflective elements in a RIS significantly impacts the performance of the UAV-RIS-assisted communication system. To investigate this, simulations were conducted to analyze how varying the number of RIS elements influences key performance metrics, including SINR, Coverage Area, and Energy Efficiency. The analysis was performed for three distinct environments: urban, suburban, and rural, reflecting different interference levels and path loss characteristics.