Unmanned Aerial Vehicles (UAVs) are an indispensable core component of low altitude economic networks. It is very critical for achieving efficient UAV image transmission of air-to-ground communication. Due to the changes in flight area and unstable channel conditions, the signal-to-noise and transmission rate change rapidly. To adapt to these changes, we propose a channel adaptive encoding and decoding method for UAV image transmission. The proposed method includes a lightweight feature extraction module, a channel-wise feature enhance module, a transmission rate adaptive module, and the corresponding decoding module. The lightweight feature extraction module can quickly extract local detailed features and long-range spatial dependencies via residual block and mobile mamba. The channel-wise feature enhance module can enhance channel wise useful features via the involution operation and the SNR adjustment block according to channel state SNRs. The transmission rate adaptive module can further adaptively adjust the size of transmission features according to the transmission rate via the rate adjustment block and the rate mask block. The extensive experimental results on the SIRI-WHU, WHU-RS19, AID, and UCMerced Land Use datasets demonstrate that our method obtains higher PSNR, MS-SSMI, LPIPS and ACC than state-of-the-art methods.

Large Artificial Intelligence Models (LAMs) possess powerful learning capabilities and are regarded as key technologies for addressing communication challenges in the future sixth-generation (6G) wireless networks. However, their massive parameters make them difficult to deploy on computation resource-constrained end nodes. Recently, large-small model collaboration has been extensively studied, but most works assume homogeneous computational resources across end nodes. This assumption neglects the heterogeneity among nodes, potentially causing significant performance degradation or even system failures due to improper resource allocation and task partitioning. To address this challenge, we propose a large-small model collaboration framework that accounts for heterogeneous computational resources and limited wireless communication bandwidth. In this proposed framework, end nodes are responsible for data collection and local inference using small models. They also cooperate with the edge server that provides large model inference and model update. We design a joint optimization strategy that considers data transmission optimization and transmission resource allocation. The primary objective of this strategy is to enhance the inference accuracy of the framework by maximizing the mean average precision (mAP). Furthermore, we derive a closed-form lower bound for the mAP of the proposed framework. Simulations based on object detection experiments demonstrate that the proposed framework significantly outperforms existing frameworks under different communication bandwidths and data scales.

A radio map captures the spatial distribution of wireless channel parameters, such as the strength of the signal received, across a geographic area. The problem of fine-grained three-dimensional (3D) radio map construction involves inferring a high-resolution radio map for the two-dimensional (2D) area at an arbitrary target height within a 3D region of interest, using radio samples collected by sensors sparsely distributed in that 3D region. Solutions to the problem are crucial for efficient spectrum management in 3D spaces, particularly for drones in the rapidly developing low-altitude economy. However, this problem is challenging due to ultra-sparse sampling, where the number of collected radio samples is far fewer than the desired resolution of the radio map to be estimated. In this paper, we design RadioLAM, a fine-grained 3D radio map construction paradigm built on generative Large Artificial Intelligence Models (LAMs). RadioLAM employs the creative power and the strong generalization capability of LAM to address the ultra-sparse sampling challenge. It consists of three key blocks: 1) an augmentation block, using the radio propagation model to project the radio samples collected at different heights to the 2D area at the target height; 2) a generation block, leveraging a diffusion-based LAM under an Mixture of Experts (MoE) architecture to generate a candidate set of fine-grained radio maps for the target 2D area; and 3) an election block, utilizing the radio propagation model as a guide to find the best map from the candidate set. Extensive simulations show that RadioLAM is able to solve the fine-grained 3D radio map construction problem efficiently from an ultra-low sampling rate of 0.1%, and significantly outperforms state-of-the-art (SOTA). Furthermore, real-world experiments also confirm that RadioLAM achieves superior performance compared to SOTA.

Radio map presents communication parameters of interest, e.g., received signal strength, across a geographical region in a specific frequency band. It can be leveraged to improve the efficiency of spectrum utilization. With the rapid development of radio technology and the proliferation of radioenabled devices, there is an increasing demand for finer granularity (higher resolution) in radio maps. However, the problem of fine-grained radio map construction is uniquely challenging, as it requires to utilize an extremely small number of radio samples collected by sparsely distributed sensor devices to infer the map. To address the challenge, we propose a hybrid device-edge-cloud collaboration paradigm called CollaboRadio. CollaboRadio first groups sensor devices into multiple clusters, with cluster locations optimized based on principles of radio propagation. Next, it leverages a small AI model on each edge server to generate a local radio map for the respective cluster region from radio samples collected by intra-cluster sensors. Finally, it employs a large AI model in the cloud to construct a global radio map for the entire region from the local maps produced by edge servers of different clusters. For the implementation of CollaboRadio, we develop an UNet-based small model for the edge server and a Transformerbased large model for the cloud. Extensive simulations show that CollaboRadio is capable of constructing fine-grained radio maps from an ultra-low sampling rate of 0.1%, and significantly outperforms state-of-the-art.

The Low-Altitude Intelligent Network (LAIN) serves as a critical enabler for low-altitude economic development, showing significant potential in areas such as environmental monitoring. These applications typically require extensive computational resources to drive large-scale models, posing challenges to resource-constrained aerial platforms like unmanned aerial vehicles (UAVs). Although air-ground model collaboration has been studied by some early studies, most approaches assume homogeneous computational resources, overlooking the heterogeneity among UAVs as end nodes. To address this challenge, we introduce an air-ground model collaboration framework that considers heterogeneous computational resources among multiple UAVs and limited wireless transmission bandwidth between UAVs and ground servers. In this framework, UAVs working as end nodes handle data collection and local inference with small models, while the edge servers on the ground perform large model inference and updates. We propose a joint optimization strategy that optimizes both data transmission and resource allocation, with the goal of improving the framework’s inference accuracy by maximizing the mean average precision (mAP). Simulations on object detection tasks show that our framework outperforms existing methods under different communication bandwidths and data scales.

Unmanned aerial vehicles (UAVs) are increasingly deployed for real-time intelligent tasks requiring large model inference, but their limited onboard computing power brings the necessity of model collaboration between the edge UAVs and cloud servers on the ground. Existing frameworks of edge-cloud collaborations often assume homogeneous data distribution among the UAVs, which fail to address real-world diversities. This paper presents a novel edge-cloud collaborative inference framework for aerial networks with distributionally diverse data. The framework introduces a joint optimization of confidence thresholds and quantization strategies to balance inference ac-curacy and bandwidth usage. We conduct probability density function (PDF) estimations on the confidence levels of different scenario datasets, which guide decisions on data uploading and quantization. Simulation results on the CIFAR-10 dataset demonstrate that our approach significantly enhances system accuracy compared to frameworks for independent identically distributed data, validating its effectiveness in diverse UAV scenarios.

A radio map captures the spatial distribution of wireless channel parameters, such as the strength of the signal received, across a geographic area. The problem of fine-grained three-dimensional (3D) radio map estimation aims to infer a high-resolution radio map for the two-dimensional (2D) area at an arbitrary target height within a 3D region of interest, using radio samples collected by sensors sparsely distributed in that 3D region. Solutions to the problem are crucial for the improvement in drone spectrum utilization in the rapidly developing low-altitude economy. However, this problem is challenging due to ultra-sparse sampling, where the number of collected radio samples is far fewer than the desired resolution of the radio map to be estimated. In this paper, we design a generative framework called GenRadio which employs the creative power of generative AI to address the ultra-sparse sampling challenge and solve the problem. GenRadio consists of three key blocks: (1) an augmentation block that uses radio propagation models to project samples collected at different heights to the 2D area at the target height; (2) a generation block that employs a diffusion model under a Mixture of Experts (MoE) architecture to generate a diverse set of fine-grained radio map candidates; and (3) an election block that leverages radio propagation models to identify the best map candidate from the diverse set. Experimental results demonstrate that GenRadio efficiently solves the fine-grained 3D radio map estimation problem from an ultra-low sampling rate of 0.1%, and significantly outperforms state-of-the-art.

Radio map presents communication parameters of interest, e.g., received signal strength, across a geographical region. It can be leveraged to improve the efficiency of spectrum utilization. With the rapid development of radio technology and the proliferation of radio-enabled devices, there is an increasing demand for finer granularity (higher resolution) in radio maps. However, the problem of fine-grained radio map construction is uniquely challenging, as it requires to utilize an extremely small number of radio samples collected by sparsely distributed sensors to infer the map. To address the problem, we propose a novel edge-cloud model collaboration paradigm. This paradigm first groups sensors into multiple clusters, with cluster locations optimized based on principles of radio propagation. Next, a small model on each edge device generates a local radio map for the respective cluster region using intra-cluster radio samples. Finally, a large model in the cloud constructs a global radio map for the entire region from the local maps produced by edge devices of different clusters. To evaluate the proposed paradigm, we also develop an UNet-based small model for the edge device and a Transformer-based large model for the cloud. Extensive simulations show that our paradigm is capable of constructing fine-grained radio maps from an ultra-low sampling rate of 0.1%, and significantly outperforms state-of-the-art.

The widespread adoption of large artificial intelligence (AI) models has enabled numerous applications of the Internet of Things (IoT). However, large AI models require substantial computational and memory resources, which exceed the capabilities of resource-constrained IoT devices. End-edge collaboration paradigm is developed to address this issue, where a small model on the end device performs inference tasks, while a large model on the edge server assists with model updates. To improve the accuracy of the inference tasks, the data generated on the end devices will be periodically uploaded to edge server to update model, and a distilled model of the updated one will be transmitted back to the end device. Subjected to the limited bandwidth for the communication link between the end device and the edge server, it is important to investigate whether the system should allocate more bandwidth to data upload or to model transmission. In this paper, we characterize the impact of data upload and model transmission on inference accuracy. Subsequently, we formulate a bandwidth allocation problem. By solving this problem, we derive an efficient optimization framework for the end-edge collaboration system. The simulation results demonstrate our framework significantly enhances mean average precision (mAP) under various bandwidths and datasizes.

The growth of the Internet of Things (IoT) has reshaped the way devices, systems, and applications connect, leading to an enormous surge in data generation across various domains. This expansion, paired with the exponential increase in IoT devices, requires advanced data analysis capabilities to manage the multimodal sensory data collected by the massive IoT devices in real time, such as sensor outputs, visual data, audios, and videos. To address this challenge, large generative artificial intelligent (AI) models are designed, showing promise in processing multimodal data. However, deploying these models on IoT devices is constrained by limited computational power, memory, and energy resources, preventing full realization of their potential for real-time IoT systems. To address these limitations, we propose an innovative end-edge collaborative model framework between end nodes and edge servers, designed to balance computational load and optimize resource use. This approach transmits both extracted features and residual mapping data from end nodes to edge servers, allowing for spectrum efficient data handling across the network. Our work formulates an optimization strategy to enhance mean average precision (mAP) by adjusting task distribution, bandwidth, and data quantization in response to real-time network and device conditions. Comprehensive simulations demonstrate the proposed approach’s superiority over conventional centralized edge model computing and distributed end model computing frameworks, achieving enhanced efficiency across various communication rates in real time.

The radio map presents communication parameters of interest, e.g., received signal strength, at every point across a geographical region. It can be leveraged to improve the efficiency of spectrum utilization in the region, particularly critical for unlicensed WiFi spectrum. The problem of fine-grained radio map estimation is to utilize radio samples collected by sensors sparsely distributed in the region to infer a high-resolution radio map. This problem is challenging due to the ultra-low sampling rate, i.e., because the number of available samples is far fewer than the high resolution required for radio map estimation. We propose WiFi-Diffusion – a novel generative framework for achieving fine-grained WiFi radio map estimation using diffusion models. WiFi-Diffusion employs the creative power of generative AI to address the ultra-low sampling rate challenge and consists of three blocks: 1) a boost block, using prior information such as the layout of obstacles to optimize the diffusion model; 2) a generation block, leveraging the diffusion model to generate a candidate set of fine-grained radio maps; and 3) an election block, utilizing the radio propagation model as a guide to find the best fine-grained radio map from the candidate set. Extensive simulations demonstrate that 1) the fine-grained radio map generated by WiFi-Diffusion is ten times better than those produced by state-of-the-art (SOTA) when they use the same ultra-low sampling rate; and 2) WiFi-Diffusion achieves comparable fine-grained radio map quality with only one-fifth of the sampling rate required by SOTA.

The future sixth-generation (6G) of wireless networks is expected to surpass its predecessors by offering ubiquitous coverage through integrated air-ground deployments in both communication and computing domains. In such networks, aerial platforms, such as unmanned aerial vehicles (UAVs), conduct artificial intelligence (AI) computations based on multi-modal data to support diverse applications including surveillance and environment construction. However, these multi-domain inference and content generation tasks require large AI models, demanding powerful computing capabilities and finely tuned inference models trained on rich datasets, thus posing significant challenges for UAVs. To tackle this problem, we propose an integrated air-ground edge-cloud model framework, in which UAVs serve as edge nodes for data collection and small model computation. Through wireless channels, UAVs collaborate with ground cloud servers providing large model computation and model updating for edge UAVs. With limited wireless communication bandwidth, the proposed framework faces the challenge of information exchange scheduling between the edge UAVs and the cloud server. To tackle this, we present joint task allocation, transmission resource allocation, transmission data quantization design, and edge model update design to enhance the inference accuracy of the integrated air-ground edge-cloud model evolution framework by mean average precision (mAP) maximization. A closed-form lower bound on the mAP of the proposed framework is derived based on the mAP of the edge model and mAP of the cloud model, and the solution to the mAP maximization problem is optimized accordingly. Simulations, based on results from vision-based classification experiments, consistently demonstrate that the mAP of the proposed integrated air-ground edge-cloud model evolution framework outperforms both a centralized cloud model framework and a distributed edge model framework across various communication bandwidths and data sizes.

Radio map presents communication parameters of interest, e.g., radio power, across a geographical region in a specific frequency band. It can be leveraged to improve the efficiency of spectrum utilization. The problem of constructing a radio map involves utilizing measurements from sparsely distributed sensors to infer the parameter of interest at every point in the region. One major challenge of solving this problem is ensuring a strong generalization capability for the solution, i.e., ensuring that the solution is able to make accurate inferences even in the presence of unknown system variables that can affect radio propagation, such as obstacle layouts and weather conditions. This paper focuses on analyzing and optimizing the generalization capability of radio map construction, an aspect that has been neglected in prior research as far as we know. We design RobUNet—a UNet-based algorithm that captures multi-scale features of radio maps. It further leverages mechanisms of residual connection, channel attention, and pixel attention to enhance its generalization capability. Simulations based on real-world dataset show that (i) the radio maps constructed by RobUNet are much more accurate than those constructed by many existing solutions in the presence of unknown system variables, and (ii) the accuracy of the radio maps constructed by RobUNet when system variables are unknown is close to that of the radio maps constructed by RobUNet when system variables are known. As a result, the generalization capability of RobUNet is much stronger than that of state-of-the-art.

Full-duplex (FD) is a promising technique to be discussed in the sixth-generation (6G) of wireless networks, which provides high spectrum efficiency (SE) of wireless transmissions, crucial for supporting emerging applications like large model computation services for cellular users. Such applications should support multiple users with massive concurrent data exchange, posing challenges for highly mobile users and those navigating complex wireless channels, such as aerial user nodes (AUNs). In this paper, we propose a FD air-ground model collaboration framework containing ground cloud node (GCN), aerial edge node (AEN), and AUNs. AUNs transmit sensory data to the AEN for model-driven feature extraction, enabling significant data compression before transmission to GCNs. AEN then uses FD transmission to upload extracted feature to GCN for large model computation tasks, such as visual target detection. To maximize accurate target detection within given timeframe, we propose joint optimization of AEN and AUN transmission power, AEN trajectory, and feature quantization. The proposed non-convex problem is decoupled into two sub-problems: transmission power and quantization design, and AEN trajectory design, and solved iteratively with mathematical analysis and convex optimizations. Simulation results show that our FD framework supports about 4 times of the target detection compared to a half-duplex (HD) framework, and about 20 times compared to a non-AEN framework.

Intelligent omni-directional surfaces (IOS), which can simultaneously reflect and refract incident signals, are considered as a promising solution for enhancing communication quality. To conduct joint beamforming of the BS and IOS, beam training is introduced such that perfect channel state information is not required anymore. However, the propagation environment is usually dynamically varying in practice, leading to frequent beam training procedures and huge training overhead. In this paper, we propose a transfer learning based beam training scheme for the IOS-assisted multi-user system to adapt to the dynamically changing propagation environment. We first build on an offline phase to train a beam prediction model that outputs the optimal beam with the highest data rate given only the received power of a small number of beams as the input. Then a transfer learning based method is developed such that the above beam prediction model can be updated to adapt to the dynamic environment rapidly. Simulation results demonstrate that the proposed scheme outperforms the existing beam training schemes in dynamic environments in terms of the convergence speed and the sum rate.

Efficiently and accurately measuring forest structure is of great significance for high-quality assessment of forest resources. Backpack laser scanning (BLS) has become a common device to acquire forest structural information due to its low cost and high time efficiency. However, complex forest environments bring challenges to BLS-based forest mapping, which faces problems with incomplete data and poor mapping accuracy. In this article, we design a disassembly-free dual-scanner BLS system for complete and accurate forest mapping. We first execute a high precision automatic self-calibration of dual laser scanners by means of angle compensation and the fixed rotation angle. Then, a simultaneous localization and mapping (SLAM) framework by combining the natural feature of trees and Inertial Measurement Unit (IMU) measurements is proposed, in which IMU provides priori motion estimation and motion compensation for dual scanners, and the natural feature of the forest is used to correct motion. The proposed method is validated in three small-scale forest plots with size of 0.1 ha. Experimental results show well performance in terms of mapping accuracy, where the mean errors and the root square mean errors are less than 3.0 cm in both horizontal and vertical directions. Our study demonstrates the effectiveness of the proposed strategy and has the potential to perform accurate and complete mapping in understory.

RRAM crossbars provide a promising hardware plat-form to accelerate matrix-vector multiplication in deep neural networks (DNNs). To exploit the efficiency of RRAM crossbars, extensive research ex-amining architecture, data representation, logic de-sign as well as device programming should be conducted. This extensive scope of research aspects is enabled and required by the versatility of RRAM cells and their organization in a computing system. These research aspects affect or benefit each other. Therefore, they should be considered systematically to achieve an efficient design in terms of design complexity and computational performance in accelerating DNNs. In this paper, we illustrate study exam-ples on these perspectives on RRAM crossbars, in-cluding data representation with pulse widths, archi-tecture improvement, implementation of logic functions using RRAM cells, and efficient programming of RRAM devices for accelerating DNNs.

Intelligent reflecting surfaces (IRSs), which are capable of adjusting radio propagation conditions by controlling the phase shifts of the waves that impinge on the surface, have been widely analyzed for enhancing the performance of wireless systems. However, the reflective properties of widely studied IRSs restrict the service coverage to only one side of the surface. In this paper, to extend the wireless coverage of communication systems, we introduce the concept of intelligent omni-surface (IOS)-assisted communication. More precisely, an IOS is an important instance of a reconfigurable intelligent surface (RIS) that can provide service coverage to mobile users (MUs) in a reflective and a refractive manner. We consider a downlink IOS-assisted communication system, where a multi-antenna small base station (SBS) and an IOS jointly perform beamforming, for improving the received power of multiple MUs on both sides of the IOS, through different reflective/refractive channels. To maximize the sum-rate, we formulate a joint IOS phase shift design and SBS beamforming optimization problem, and propose an iterative algorithm to efficiently solve the resulting non-convex program. Both theoretical analysis and simulation results show that an IOS significantly extends the service coverage of the SBS when compared to an IRS.

Deep neural networks have successfully been applied to various fields. The efficient deployment of neural network models emerges as a new challenge. Processing-in-memory (PIM) engines that carry out computation within memory structures are widely studied for improving computation efficiency and data communication speed. In particular, resistive memory crossbars can naturally realize the dot-product operations and show great potential in PIM design. The common practice of a current-based design is to map a matrix to a crossbar, apply the input data from one side of the crossbar, and extract the accumulated currents as the computation results at the orthogonal direction. In this study, we propose a novel PIM design concept that is based on the crossbar connections. Our analysis on star-mesh network transformation reveals that in a crossbar storing both input data and weight matrix, the dot-product result is embedded within the network connection. Our proposed connection-based PIM design leverages this feature and discovers the latent dot-products directly from the connection information. Moreover, in the connection-based PIM design, the output current range of resistive crossbars can easily be adjusted, leading to more linear conversion to voltage values, and the output circuitry can be shared by multiple resistive crossbars. The simulation results show that our design can achieve on average 46.23% and 33.11% reductions in area and energy consumption, with a merely 3.85% latency overhead compared with current-based designs.

Memristor-based crossbars are considered to be promising candidates to accelerate vector-matrix computation in deep neural networks. Before being applied for inference, mem-ristors in the crossbars should be programmed to conductances corresponding to the network weights after software training. Existing programming methods, however, adjust conductances of memristors individually with many programming-reading cycles. In this paper, we propose an efficient programming framework for memristor crossbars, where the programming process is partitioned into the predictive phase and the fine-tuning phase. In the predictive phase, multiple memristors are programmed simultaneously with a memristor programming model and IR-drop estimation. To deal with the programming inaccuracy resulting from process variations, noise and IR-drop and move conductances to target values, memristors are fine-tuned afterwards to reach a specified programming accuracy. Simulation results demonstrate that the proposed method can reduce the number of programming-reading cycles by up to 94.77% and 90.61% compared to existing one-by-one and row-by-row programming methods, respectively.

In this paper we propose a cluster-based platoon handoff protocol (CPHP), in which the platoon is divided into clusters to minimize its handoff delay between two adjacent cells. Via the vehicle-to-vehicle (V2V) communications, multiple clusters are constructed within one platoon and only the cluster head (CH) communicates with the base station (BS) via the vehicle-to-infrastructure (V2I) communications. In this way, the number of V2I links can be greatly saved and the signaling overhead is reduced, thereby shortening the handoff delay. Specifically, we formulate a delay minimization problem based on a discrete-time Markov chain and then optimize the number of clusters and spectrum resource allocation. Simulation results demonstrate a significant decrease in the handoff delay of the platoon. The optimal expected delay under a platoon with 50 vehicles by utilizing the proposed CPHP is 15% smaller than that of the traditional handoff scheme.

In-memory computing has been applied in different fields due to its superior speed and energy efficiency. Among a variety of memory technologies that have been explored, resistive memory has widely been adopted for various purposes, including Processing-In-Memory (PIM) for neural networks and Logic-In-Memory (LIM) for general logic operations. PIM has intensively been studied in recent years, while the progress in developing LIM computing falls behind. LIM computing is usually implemented based on MAGIC operations, which require inputs to be aligned regularly along rows or columns in a memory crossbar. As the intermediate data generated during the logic execution are normally scattered across the memory crossbar, alignment operations are inserted to align the data, which often costs numerous cycles and dominates the overall latency. In current MAGIC-based designs, alignment operations induce a significant overhead in either area or latency. Therefore, the Area-Latency-Product (ALP), known as a key metric for circuit performance, still has significant optimization potential in LIM computing. In this work, we leverage peripheral circuitry to conduct alignment operations and propose a novel mapping framework to optimize the latency and area costs. Intermediate data are read out, processed in peripheral circuits, then in parallel written back into target cells of the memory crossbar. The approach eliminates the use of redundant memory cells, leading to area reduction. Moreover, it enables simultaneous alignments of multiple intermediate data, which can decrease the overall latency significantly. Based on simulation results, our proposed mapping framework can achieve around 93% ALP reductions on average compared with prior designs with merely 2.13% total area overhead.

Processing-in-memory engines have successfully been applied to accelerate deep neural networks. For improving computing efficiency, spiking-based designs are widely explored. However, spiking-based designs quantize inter-layer signals naturally, leading to performance loss. In addition, the spike mismatch effect makes digital processing necessary, impeding direct signal transfer between layers and thus resulting in longer latency. In this paper, we propose a novel neuron design based on pulse width modulation, avoiding the quantization step and bypassing spike mismatch via the continuous activation. The computation latency and circuit complexity can significantly be reduced due to the absence of quantization and digital processing steps, while keeping a competitive performance. Simulation results show that the proposed neuron design can achieve > 100× speedup compared with spiking-based designs. The area and power consumption can be reduced up to 74.87% and 25.63%.

RRAM-based crossbars are a promising hardware platform to accelerate computations in neural networks. Before such a crossbar can be used as an accelerator for neural networks, RRAM cells should be programmed to target resistances to represent weights in neural networks. However, this process degrades the valid range of the resistances of RRAM cells from the fresh state, called aging effect. Therefore, after a certain number of programming iterations, these RRAM cells cannot be programmed reliably anymore, affecting the classification accuracy of neural networks negatively. In addition, process variations during manufacturing and noise during programming of RRAM cells also lead to significant accuracy degradation. To solve the problems described above, in this paper, we introduce a software/hardware codesign framework to reduce the aging effect in RRAM crossbars. To counter process variations and noise, we first model them as random variables and then modify the computations in software training considering these variables. Simulation results show that the lifetime of RRAM crossbars can be extended by up to 11 times with the codesign framework and the mean value and the standard deviation of the inference accuracy under process variations and noise can be improved significantly.

In this paper, we consider the cellular Internet of unmanned aerial vehicles (UAVs), where UAVs sense data for multiple tasks and transmit the data to the base station (BS). To quantify the “freshness” of the data at the BS, we bring in the concept of the age of information (AoI). The AoI is determined by the time for UAV sensing and that for UAV transmission, and gives rise to a trade-off within a given period. To minimize the AoI, we formulate a joint sensing time, transmission time, UAV trajectory, and task scheduling optimization problem. To solve this problem, we first propose an iterative algorithm to optimize the sensing time, transmission time, and UAV trajectory for completing a specific task. Afterwards, we design the order in which the UAV performs data updates for multiple sensing tasks. The convergence and complexity of the proposed algorithm, together with the trade-off between UAV sensing and UAV transmission, are analyzed. Simulation results verify the effectiveness of our proposed algorithm.

In this paper, we consider the cellular Internet of unmanned aerial vehicles (UAVs), where UAVs sense data with onboard sensors for multiple sensing tasks and transmit the data to the base station (BS). To quantify the “freshness” of the data at the BS, we bring in the concept of the age of information (AoI). The AoI is determined by the time for UAV sensing and that for UAV transmission, which gives rise to a trade-off within a given period. To minimize the AoI, we formulate a joint sensing time, transmission time, UAV trajectory, and task scheduling optimization problem. This NP-hard problem can be decoupled into two subproblems. We first propose an iterative algorithm to optimize the sensing time, transmission time, and UAV velocity for completing a specific task. Afterwards, we design the order in which the UAV performs data updates for multiple sensing tasks. The convergence and complexity of the proposed algorithm, together with the trade-off between UAV sensing and UAV transmission, are analyzed. Simulation results show that the AoI with the proposed algorithm is about 15% lower than that of the greedy algorithm, and over 40% lower than that of the random algorithm.

Memristor-based crossbars have been applied successfully to accelerate vector-matrix computations in deep neural networks. During the training process of neural networks, the conductances of the memristors in the crossbars must be updated repetitively. However, memristors can only be programmed reliably for a given number of times. Afterwards, the working ranges of the memristors deviate from the fresh state. As a result, the weights of the corresponding neural networks cannot be implemented correctly and the classification accuracy drops significantly. This phenomenon is called aging, and it limits the lifetime of memristor-based crossbars. In this paper, we propose a co-optimization framework combining software training and hardware mapping to reduce the aging effect. Experimental results demonstrate that the proposed framework can extend the lifetime of such crossbars up to 11 times, while the expected accuracy of classification is maintained.

In this paper, we consider a cellular controlled unmanned aerial vehicle (UAV) network in which multiple UAVs cooperatively complete each sensing task. We first propose a sense-and-send protocol where the UAVs collect sensory data of the tasks and transmit the collected data to the base station. We then formulate a joint trajectory, sensing location, and UAV scheduling optimization problem that minimizes the completion time for all the sensing tasks in the network. To solve this NP-hard problem efficiently, we decouple it into three subproblems: 1) trajectory optimization; 2) sensing location optimization; and 3) UAV scheduling. An iterative trajectory, sensing, and scheduling optimization (ITSSO) algorithm is proposed to solve these subproblems jointly. The convergence and complexity of the ITSSO algorithm, together with the system performance are analyzed. Simulation results show that the proposed ITSSO algorithm saves the task completion time by 15% compared to the noncooperative scheme.

In this paper, we consider a single-cell cellular network with a number of cellular users (CUs) and unmanned aerial vehicles (UAVs), in which multiple UAVs upload their collected data to the base station (BS). Two transmission modes are considered to support the multi-UAV communications, i.e., UAV-to-network (U2N) and UAV-to-UAV (U2U) communications. Specifically, the UAV with a high signal-to-noise ratio (SNR) for the U2N link uploads its collected data directly to the BS through U2N communication, while the UAV with a low SNR for the U2N link can transmit data to a nearby UAV through underlaying U2U communication for the sake of quality of service. We first propose a cooperative UAV sense-and-send protocol to enable the UAV-to-X communications, and then formulate the subchannel allocation and UAV speed optimization problem to maximize the uplink sum-rate. To solve this NP-hard problem efficiently, we decouple it into three sub-problems: U2N and cellular user (CU) subchannel allocation, U2U subchannel allocation, and UAV speed optimization. An iterative subchannel allocation and speed optimization algorithm (ISASOA) is proposed to solve these sub-problems jointly. The simulation results show that the proposed ISASOA can upload 10% more data than the greedy algorithm.

In this paper, we consider a single-cell cellular network with a number of cellular users (CUs) and unmanned aerial vehicles (UAVs), in which multiple UAVs upload their collected data to the base station (BS). Two communication modes are considered to support the multi-UAV communications, i.e., UAV-to-infrastructure (U2I) and UAV-to-UAV (U2U) communications. We then formulate the subcarrier allocation and trajectory design problem to maximize the uplink sum-rate taking the delay of sensing tasks into consideration. To solve this NP-hard problem efficiently, we decouple it into three sub-problems: U2I and cellular user (CU) subcarrier allocation, U2U subcarrier allocation, and UAV trajectory design. An iterative subcarrier allocation and trajectory design algorithm (ISATCA) is proposed to solve these sub-problems jointly. Simulation results show that the proposed ISATCA can upload 20% more data than the one without U2U communication.

In this paper, we consider a cellular controlled unmanned aerial vehicle (UAV) sensing network in which multiple UAVs cooperatively complete each sensing task. We formulate a joint trajectory, sensing location, and UAV scheduling optimization problem that minimizes the completion time for all the sensing tasks in the network. To solve this NP-hard problem efficiently, we decouple it into three sub-problems: trajectory optimization, sensing location optimization, and UAV scheduling. An iterative trajectory, sensing, and scheduling optimization (ITSSO) algorithm is proposed to solve these sub-problems jointly. The convergence of the proposed algorithm and the dominated factors on the system performance are analysed. Simulation results show that the task completion time obtained by the proposed ITSSO algorithm is 15% less than that by non- cooperative scheme.

In this paper, we study the resource allocation problem for a single-cell non-orthogonal multiple access (NOMA) relay network where an OFDM amplify-and-forward relay allocates the spectrum and power resources to the source-destination (SD) pairs. We aim to optimize the resource allocation to maximize the average sum-rate. The optimal approach requires an exhaustive search, leading to an NP-hard problem. To solve this problem, we propose two efficient many-to-many two-sided SD pair-subchannel matching algorithms, in which the SD pairs and sub-channels are considered as two sets of players chasing their own interests. The proposed algorithms can provide a sub-optimal solution to this resource allocation problem in affordable time. Both the static matching algorithm and the dynamic matching algorithm converge to a pair-wise stable matching after a limited number of iterations. Simulation results show that the capacity of both proposed algorithms in the NOMA scheme significantly outperforms the conventional orthogonal multiple access scheme. The proposed matching algorithms in NOMA scheme also achieve a better user-fairness performance than the conventional orthogonal multiple access.

In this paper, we study the resource allocation problem for a single-cell non-orthogonal multiple access (NOMA) relay network where an OFDM amplify-and-forward (AF) relay allocates the spectrum and power resources to the source-destination (SD) pairs. We aim to optimize the spectrum and power resource allocation to maximize the total sum-rate. This is a very complicated problem and the optimal approach requires an exhaustive search, leading to a NP hard problem. To solve this problem, we propose an efficient many-to-many two sided SD pair-subchannel matching algorithm in which the SD pairs and sub-channels are considered as two sets of rational and selfish players chasing their own interests. The algorithm converges to a pair-wise stable matching after a limited number of iterations with a low complexity compared with the optimal solution. Simulation results show that the sum-rate of the proposed algorithm approaches the performance of the optimal exhaustive search and significantly outperforms the conventional orthogonal multiple access scheme, in terms of the total sum-rate and number of accessed SD pairs.

In this paper, we study the resource allocation problem in a small cell Orthogonal Frequency Division Multiple Access (OFDMA) network consisting of multiple access points (APs) and subscribed users. We formulate the subchannel allocation and the user assignment as a joint optimization problem, and solve it utilizing a novel three-sided matching algorithm. In the proposed algorithm, a unified cooperative framework is constructed in which we discuss both the non-overlapping and the overlapping coalitions formed by the APs, as well as the traditional non-cooperative case.