Deploying artificial intelligence (AI) algorithms at the network edge is essential for enabling next-generation network applications. In edge inference systems, features are extracted from distributed sensors and transmitted to an edge server for remote inference. However, traditional communication systems are primarily designed for reliable bit-level transmission and do not account for the unique characteristics of AI-based inference. A key property of AI inference is robustness: the ability to tolerate distortions in input features without significantly degrading performance. Leveraging this property, we propose a robustness-aware framework for energy-efficient edge inference that reduces energy consumption during feature transmission. By allowing nonzero bit error rates (BERs) in the transmitted features, our approach exploits the inherent robustness of AI models to reduce energy consumption while preserving inference accuracy. We begin by quantifying the robustness of inference using the classification margin and analyzing the relationship between classification accuracy and the average BER across distributed sensors. Based on this analysis, we formulate an optimization problem that seeks to minimize the total energy consumption of the sensors while satisfying classification accuracy requirements under radio resource constraints. To solve this problem, we decompose it into subproblems involving sensor selection, power control, and bandwidth allocation. These are efficiently addressed using a low-complexity alternating iterative algorithm. Simulation results demonstrate that the proposed robustness-aware inference framework can effectively tolerate bit errors during feature transmission, achieving significant energy reductions while maintaining target classification accuracy. Our design consistently outperforms benchmark schemes.

Ultraviolet (UV) communication is a promising technology for civilian and military secure communication systems due to nonline-of-sight transmission, low background noise, and high local security. The full-duplex relay-assisted UV communications can achieve longer communication distances and higher efficiency of time–frequency utilization compared with direct UV communications. However, due to the strong scattering effect, serious interrelay-interference (IRI) is inevitably introduced in multihop full-duplex relay links. To mitigate the impacts of IRI, this article proposes an alternate iterative-Newton method (AINM) to optimize jointly the relay placement and transmit powers of each relay. We further propose a space-division coupled full-duplex relay configuration to reduce the influence of IRI. Numerical results show that the proposed AINM can significantly decrease the bit-error rate (BER); and the proposed space-division scheme can further decrease the BER but sacrifices about half achievable information rate. Besides, we demonstrate that, when the communication links are strong, the distance between adjacent relays should gradually decrease and the transmit power increase from the source node to the destination node. However, when the communication links are weak, each relay should adopt its maximum transmit power to achieve the minimum BER.

The strong scattering effects of ultraviolet (UV) signals in the atmosphere enables the non-line-of-sight (NLOS) ability of UV communications, but introduces large time dispersion of the channel impulse response (CIR), which inevitably results in serious inter-symbol interferences (ISIs). Because a larger transmitting rate can introduce a larger ISI, it is meaningful to study the impact of CIR on the ISI and further quantify the achievable information rate (AIR) of NLOS UV communications. In this paper, we analyzed the influence of both coplanar and non-coplanar geometries on the CIR for multiple scattering effects and proposed a modified Gamma function (MGF) to precisely characterize the CIR. Based on the MGF model, we quantified the ISIs and derived the bit-error rate (BER) and the AIR under thermal noise for both on-off keying (OOK) modulation and 2-pulse-position modulation (2-PPM) using photon-counting receivers. The simulation results showed that the proposed MGF model can achieve higher fitting precision of both CIRs and BERs compared with the existing Gamma function model. Besides, we demonstrated that when the thermal noise photons are less than 10, the OOK and 2-PPM can achieve an AIR of 15 Mbit/s and 13 Mbit/s, respectively, under typical system geometries.

Optical positioning techniques have inherent ad-vantages such as high precision, low power consuming and immunity to the electromagnetic interference. However, it is challenging to estimate the non-line-of-sight (NLOS) targets using current optical positioning methods. In this work, we propose an NLOS ultraviolet positioning method based on a simplified single-scattering channel model, which can obtain both the location and the pointing direction of the transmit-ter. Numerical results demonstrated that the proposed NLOS ultraviolet positioning method can achieve a positioning error less than 2 meters and an azimuth error less than 2 degrees by using two photon-counting receivers under typical transceiver geometries within 100 meters. Besides, we found that the positioning performance can be improved by increasing the gap between two receiving elevation angles.

The full-duplex optical communications can be achieved simply by separating the transmitting link and the receiving link in the space. However, it is challenging to achieve the non-line-of-sight (NLOS) full-duplex ultraviolet (UV) communication due to the serious self-interference caused by the strong multiple scattering effects of UV signals. To explore the capacity of NLOS full-duplex UV communications, in this paper, we first quantify the self-interference using an analytical channel impulse response function. Based on the quantified self-interference, we then derive the error rate and the corresponding achievable information rate (AIR) for on- off keying modulation, 4-digital-pulse-interval modulation and 4-pulse-position modulation. We further propose a self-interference cancelling (SIC) method to mitigate the impacts of self-interferences. Simulation results show that the proposed SIC method can significantly improve the error rate and AIR performances. Besides, we find that the NLOS full-duplex UV communication will gradually lose its advantage over the NLOS half-duplex UV communication as either the communication distance or the elevation angle increases. However, using the proposed SIC method, the NLOS full-duplex UV communication can hold its advantage in a wide range of system geometries.