Accurate and reliable 3D object detection is crucial for autonomous driving, normally achieved using camera-only or camera-LiDAR fusion methods based on BEV (Bird’s Eye View) perspective. However, visual perception through cameras alone faces significant challenges, such as ambiguous depth estimation and poor performance in low-light, while camera-LiDAR fusion, though robust, are expensive on deployment and have limitations in dust and fog weather conditions. To address these issues, we propose RAFDet (Radar-Assisted Fusion Detection), a novel multi-modality fusion framework integrating camera and radar data for enhanced 3D detection. RAFDet utilizes radar RF (Radio Frequency) images to enrich spatial details, boosting fusion effects, while radar point clouds provide precise depth information to rectify visual depth predictions. In addition, the Edge-Awareness Feature Enhancement mechanism compensates for the sparsity of radar points, further refining depth estimation and detection accuracy. Our extensive experiments demonstrate the superiority of RAFDet over current methods, highlighting its potential for autonomous driving systems. Key results show significant improvements in depth prediction and overall performance metrics, validating the effectiveness of our approach. Our data collection and testing in real traffic scenarios also reflects the robustness, accuracy, and cost-effectiveness of RAFDet for 3D object detection in autonomous driving.

Cooperative perception via Vehicle-to-Everything (V2X) wireless communication technology shares complementary perception information among autonomous vehicles and infrastructure, significantly enhancing perception capabilities by addressing limitations such as restricted range and occlusions inherent in on-board sensing. However, the limited bandwidth of V2X communication poses challenges for real-time transmission of the large data volumes required for early-and mid-level cooperative perception, necessitating consideration of communication efficiency alongside perception performance. To overcome this challenge, we propose a novel communication-efficient cooperative perception framework that achieves a better trade-off between perception performance and communication cost. By exploiting the static deployment of infrastructure sensors and the heterogeneity of the perception region, the framework performs high-precision foreground segmentation on static infrastructure-side point clouds. A feature importance map is then constructed to extract sparse yet critical features for sharing, significantly reducing communication overhead. Additionally, we design a Multi-scale Weighted Attention Feature Fusion (MWAF) module to effectively integrate vehicle-and infrastructure-side features. The experimental results demonstrate that the proposed framework outperforms existing cooperative perception methods in terms of perception performance, while reducing communication traffic by approximately 60 times and 90 times on the DAIR-V2X and DZGSet dataset, respectively, without considering additional feature compression.

Departing from traditional data security-oriented designs, the aim of anonymity is to conceal the transmitters’ identities during communications to all possible receivers. In this work, joint anonymous transceiver design at the physical (PHY) layer is investigated. We first present sender detection error rate (DER) performance analysis, where closed-form expression of DER is derived for a generic precoding scheme applied at the transmitter side. Based on the tight DER expression, a fully DER-tunable anonymous transceiver design is demonstrated. An alias channel-based combiner is first proposed, which helps the receiver find a Euclidean space that is close to the propagation channel of the received signal for high quality reception, but does not rely on the recognition of the real sender’s channel. Then, two novel anonymous precoders are proposed under a given DER requirement, one being able to provide full multiplexing performance, and the other flexibly adjusting the number of multiplexing streams with further consideration of the receive-reliability. Simulation demonstrates that the proposed joint transceiver design can always guarantee the subscribed DER performance, while well striking the trade-off among the multiplexing, diversity and anonymity performance.

LiDAR semantic segmentation plays an important role in 3D scene understanding for autonomous driving. However, the performance based on the LiDAR equipped on a vehicle may be limited due to small perception perspective, object occlusion, and sparsity of point clouds in the distance. To address these challenges, we propose a vehicle-infrastructure cooperative semantic segmentation (VICSS) framework to enhance the vehicle-side perception capability. An infrastructure feature extraction (IFE) module is employed to extract features from the roadside LiDAR point cloud. A local feature extraction (LFE) module and a global feature extraction (GFE) module are applied to the vehicle point cloud to capture the local and global features, respectively. The features from both point clouds are fused through a feature aggregation (FA) module, which applies the cross-attention mechanism to learn beneficial information from the infrastructure features. Using a dataset generated by CARLA we show that the proposed VICSS framework achieves good performance in terms of semantic segmentation accuracy.

Deep reinforcement learning (DRL) can be used to solve decision-making problems in changing and complex environments. It is widely envisioned to have great potential for enabling autonomous driving. However, the large and continuous state and action spaces of autonomous driving tasks in general lead to low exploration efficiency and significantly affect the model training speed. In this paper, we propose a novel autonomous driving framework based on DRL with an intervention module. The module utilizes historical information to predict the status of future states so that evaluation of potential actions can be inferred. By this means, an intrinsic reward is generated to prevent DRL agent from entering poor states with useless exploration. Integrating our DRL model with a perception module which utilizes supervised multi-task learning to extract useful features from raw sensor data, the proposed framework can notably improve the efficiency of DRL training. A case study on the open racing car simulator (TORCS), with inputs from both distance sensors and vision sensors, is carried out to demonstrate the effectiveness of our method.

In recent years, with the rapid development of the driverless technology, how to accurately track the target vehicle has become a concern. With the advancement of the global satellite navigation system, the positioning accuracy of the vehicle has been greatly improved. However, in some cases, GPS signal is easy to be interfered, such as in cities with high-rise buildings. Therefore, how to ensure the accuracy of positioning in complex clutter environment and achieve reliable target tracking has become the main research content of this work. In this paper, a framework suitable for vehicle-road data fusion is proposed, which mainly receives perception information from the roadside and on-board GPS. After that, based on this framework, an improved IMM-JPDA-PFF algorithm for updating the fusion is proposed to improve the accuracy of vehicle tracking in complex clutter environments. This paper evaluates the method by means of matlab simulation, and the experiments show that the method can adapt to the clutter scene with changing GPS accuracy. Besides, the adaptability and the calculation efficiency are significantly enhanced than the traditional particle filter algorithm does. Therefore, this method helps to improve the safety and reliability of automatic driving assistance systems.

Autonomous valet parking (AVP) is a popular application scenario for autonomous driving in the future. For AVP path planning, the original hybrid A-star ($\mathrm{A}^{*}$) algorithm has problems of large search costs, searching towards wrong directions and generating unreasonable parking paths. To solve these problems and generate a better path, a path planning method is proposed for typical AVP scenarios. The method divides path planning into global part and local part. The global path is planned based on graph search and state lattice algorithm. Then the hybrid $\mathrm{A}^{*}$ algorithm and Reeds-Shepp curve are modified to complete the local path planning, and finally a complete path that can be executed by the vehicle is generated. Then, a controller for path tracking based on model predictive control (MPC) is designed to overcome the shortcomings of traditional proportional integral derivative (PID) control such as overshoot and difficulty in precise control. Finally, the feasibility of the path planning and tracking method is verified by simulation using MATLAB and the vehicle simulation software CarSim. The results show that the planning efficiency and rationality are improved by implementing the proposed method, and the parking process can be done well with a small tracking error.

In the era of e-Health, privacy protection has become imperative in applications that carry personal and sensitive data. Departing from the data-perturbation based privacy-preserving techniques that reduce the fidelity of the disclosed data, in this paper we investigate anonymous communications, which mask the identity of the data sender while providing high data reliability. Focusing on the physical (PHY) layer, we first explore the break of privacy through a statistical attribute based sender detection (SD) from the receiver. Compared to the existing literature, this enables a much enhanced SD performance, especially when the users are equipped with different numbers of antennas. To counteract the advanced SD approach above, we formulate explicit anonymity constraints for the design of the anonymous precoder, which mask the sender’s PHY attributes that can be exploited by SD, while at the same time preserving the reliability of the data. Then, anonymity entropy-oriented precoders are proposed for different antenna configurations at the users, which adaptively construct a maximum number of aliases while obeying users’ signal-to-noise-ratio requirements for data accuracy. Simulation results demonstrate that the proposed anonymous precoders provide the highest level of anonymity entropy over the benchmarks, while achieving reasonable symbol error rate for the communication signal.

In this study, a Human-Lead-Platoon CACC ((HLP-CACC) controller is proposed for connected and automated vehicles to “include” human drivers in platooning process. The goal is to form a platoon between automated vehicles and human drivers so that turbulence caused by human drivers could be smoothed out by automated vehicles. Unlike the conventional CACC where only longitudinal control is automated, the proposed HLP-CACC regulates both longitudinally and laterally. In other words, the followers in an HLP-CACC platoon are fully autonomous. The controller is formulated utilizing model predictive control (MPC) solved by Chang-Hu’s method. The technology has the following advantages: 1) take advantage of human drivers’ perception to enable conditional full autonomy; 2) accommodate actuator delay in system dynamics to improve actuator control accuracy; 3) automates both longitudinally and laterally; and 4) ensures string stability in partially connected and automated vehicles environment. Both simulation tests and field tests were conducted to verify the effectiveness of the proposed algorithm. Four scenarios, including straight cruising, lane changing, U-turn and circling were tested. Sensitivity analysis was conducted for speed, turning radius, communication delay and oscillation acceleration. The results confirm that the proposed CACC controller is ready for field implementation. The computation time of the proposed optimal control is approximately$4~\sim ~8$milliseconds when running on an NVIDIA Drive PX 2 computer. Under the control of the proposed HLP-CACC, maximum longitudinal error and lateral error are both within 40 cm.

3D object detection in point cloud data is an important aspect of computer vision systems, especially for autonomous driving applications. Recent literature suggests two methods of point cloud encoders; grid-based methods tend to be fast but sacrifice accuracy, while point-based methods that are learned from raw data are more accurate, but slower. In this work, we present a novel and real-time two-stage 3D object detection framework, named PointPillars-RCNN (PP-RCNN). In the first stage, we use pillars network to encode the point cloud and generate high-qulaity 3D proposals. Benefiting from the pillars network, our framework realizes real-time detection. In the second stage, we use the Point-Pillars Feature Set Abstraction (PPSA) module to extract the point-based features from raw point cloud and pillars features, and then we use the RoI-grid feature abstraction for proposals refinement. All our detection pipelines are trained end-to-end. Extensive experiments on the KITTI benchmark shows that our approach has better performance than the one-stage PointPillars algorithm, and faster than current two-stage state-of-the-art algorithms.

This paper investigates efficient data delivery methods, through integrating non-orthogonal transmission with network coding techniques, in wireless networks with multiple sources, multiple relays, and multiple destinations. We propose properly scheduling terminals into transmitting clusters to optimize system error performance. Considering communication over general Nakagami-m fading channels, an error probability approximation analysis is first presented. To reduce computation complexity for solving the node scheduling design problem, we further propose transforming the problem into a data classification problem and applying data-driven machine learning tools to reach rapid decision-making. Taking two-user clustering as an example, simulation results show that our method can attain similar performance as the optimal solution, and significantly outperform conventional methods that orthogonalize or randomly schedule terminals. The proposed method and the analytical framework can be extended to larger networks with diverse node clusters and more complex signal propagation environments.

Intelligent connected vehicle (ICV) has been widely deemed as the key to reduce road accident rate and improve traffic efficiency. However, ensuring high communication reliability and low transmission delay in vehicular networks is challenging, especially in large-scale dynamic networks with diverse heterogeneous data exchange demands. In this paper, we investigate the potential of applying the deep reinforcement learning (DRL) technique to facilitate efficient transmission design in a class of complex multi-user vehicle-to-vehicle (V2V) networks, where conventional mathematical tools confront difficulties in solving the design optimization problems. The considered network contains several pairs of V2V links sharing the channel resource. Each link desires to communicate two types of delay-sensitive messages to support different safety-related applications with the maximum energy efficiency. We propose transforming the power/rate control problem into a Markov decision process and then solving it using the deep deterministic policy gradient (DDPG) algorithm. Simulation results show that in a two-user network our DRL-aided solution can achieve better performance than that with Lyapunov optimization. Extending the former to work in a larger network is straightforward, but it is not the case for the latter. The advantages of applying DRL to support wireless system design are thus demonstrated.

Cognitive radio (CR) has been proposed to mitigate the spectrum scarcity issue to support heavy wireless services on sub‐3GHz. Recently, broadband spectrum sensing becomes a hot topic with the help of compressive sensing technology, which will reduce the high‐speed sampling rate requirement of analog‐to‐digital converter. This paper considers sequential compressive spectrum sensing, where the temporal correlation information between neighboring compressive sensing data will be exploited. Different from conventional compressive sensing, the previous compressive sensing data will be fused into prior knowledge in current spectrum estimation. The simulation results show that the proposed scheme can achieve 98.7% detection probability under 3.5% false alarm probability and performs the best compared with the typical BPDN and OMP schemes.

This letter investigates the performance of energy detector with multiple antennas in cognitive vehicular networks, where the sensing signals of the secondary user are temporally correlated. A novel analytical method, based on new results of confidence interval estimation for auto-correlated matrix in random matrix theory, is proposed to estimate the miss detection probability. The estimate accuracy is verified via simulations. Based on such results, the impacts of signal temporal correlation on sensing performance and decision threshold are analyzed.

This paper investigates efficient transmission design in a class of multi-user cooperation networks, in which multiple information sources intend to distribute their messages to a sufficient proportion of ambient destinations with the assistance of multiple relays. We apply a relaying scheme that combines non-orthogonal transmission with network coding techniques to balance channel consumption and inter-user interference. The sources and relays are divided into clusters, terminals within each of which are allowed to non-orthogonally access the same channel. A class of finite-field network codes are adopted in the relays. We provide the methods to derive the system transmission error probability and diversity-multiplexing tradeoff (DMT), over Nakagami-m fading channels. Through error probability and finite-SNR DMT analysis for certain clustering strategies, and infinite-SNR DMT analysis for general situations, we show that the considered relaying scheme can notably improve system performance over the conventional approach that demands only orthogonal transmission in network-coded cooperation networks.

Vehicular networking is potentially an effective solution to the problems in today's transportation system. However, realizing efficient communications among vehicles with satisfactory reliability and latency requirements is challenging, especially when diverse applications are taken into consideration. In this paper, we investigate a cross-layer energy- efficient transmission design for a class of vehicular communication networks, in which two pairs of vehicle-to-vehicle links non-orthogonally share the available spectrum. Each link desires to deliver two types of messages that can support different delay-limited applications. The periodically- generated heartbeat messages should be transmitted subject to a reliability requirement, and the randomly-appeared sensing messages should be delivered with finite latency. We propose a power control strategy to achieve high energy efficiency, while ensuring the expected quality-of-service requirements, based on both channel state information in the physical layer and queue state information in the media access control layer. Simulation results show that our proposed method notably outperforms conventional methods.

This paper investigates the communication problem in a class of multi-user dual-hop networks in which multiple source terminals desire to distribute their independent messages to multiple destinations through the assistance of multiple relay terminals. We consider an efficient transmission strategy that combines network coding and non-orthogonal transmission techniques to balance the achievability of spatial diversity and channel utilization. Specifically, in addition to applying a class of finite-field network codes in the relays, we divide the sources and the relays into clusters such that terminals within each cluster can access the same channel resource. The achievable error performance under Nakagami-m fading environment is derived and is shown to significantly outperform the conventional transmission methods.

Broadcasting of periodic messages is critical for intelligent transportation systems (ITS). In this paper, we analyze the performance of the delivery of periodic messages via stochastic geometry and the technique of Padá approximation (PA) under Weibull-lognormal composite fading channels in highway scenarios. The locations of the vehicles are assumed to follow a one-dimensional homogeneous Poisson Point Process (PPP), and the slotted ALOHA MAC scheme is employed. The closed-form expressions of coverage probability, mean number of covered vehicles, and mean packet reception rate (PRR) are presented. Monte-Carlo simulations validate the accuracy of our analytical results.

This paper studies the resource allocation (RA) problem when the in- band device-to-device (D2D) technology is applied to support vehicle-to-vehicle (V2V) communications. Conventional D2D RA normally demands a sufficient level of channel knowledge to reach the optimal performance. But in vehicular communication environments, this would require tremendous signalling overhead and the acquired channel knowledge is easily outdated. Therefore, RA relying only on geographic information is more feasible. To this end, we propose a novel location-partition-based RA scheme. We first divide the cell coverage area and road into small zones, the geographic information of which is stored in a database. Satisfying the requirement that the interference generated by all V2V links to the reused cellular user (CU) is below a certain threshold, an interference matrix that reflects the interference from nodes in cell zones to road zones is established. Three types of power control methods are adopted to maximize the minimum achievable rate of the V2V links. A series of simulations are conducted to verify the performance of our proposed RA solution. The results show that our method can improve the minimum achievable rate compared with conventional location-based RA methods. The impact of different system parameters are also carefully analyzed.

This is the author accepted manuscript. The final version is available from IEEE via the DOI in this record

We investigate the achievable sum degrees of freedom (DoF) in a class of single-antenna multi-user multi-hop relay networks. The networks consist of multiple information sources and destinations, without direct signal propagation link between them, so that multiple layers of relays are deployed to assist in information delivery. We consider the situation that relays are unable to shield their receptions from the harmful self-interference and from the interference generated by other relays. Hence, ideal full-duplex relaying is not applicable. Utilizing half-duplex decode-and-forward relays, a cluster successive relaying (CSR) transmission scheme is adopted to conduct message transmission. The CSR scheme divides each layer of relays into two successively activated relay clusters to compensate the extra channel consumption demanded by the half-duplex operation. We propose two interference alignment strategies to deal with the interference issues. By properly clustering the relays in each layer, we find the asymptotically achievable sum DoF, subject to time-varying and frequency-selective fading, respectively. These results can lead to new lower bounds for the available DoF in the considered class of multi-user multi-hop networks.

In this paper, we study linear transceiver designs for indoor visible light communications (VLCs) with multiple light emitting diodes (LEDs). Specifically, we investigate VLCs including white emitting diodes and VLCs including red/green/blue (RGB) LEDs. The transmitter precoding and the offset are jointly designed by considering certain key practical lighting constraints, such as optical power, non-negativeness, and color illumination. Various non-convex transceiver design problems are formulated aiming to minimize total mean-square-error to improve transmission reliability. We show that for multi-input single-output white VLCs, the optimal precoding reduces to a simple LED selection strategy. For multi-input multi-output (MIMO) white VLCs, we prove that the optimization problem with multiple constraints can be equivalently simplified to a problem with single constraint, which enables us to propose efficient algorithms to search local optimal solutions. For MIMO RGB VLCs, by using certain useful transformations, we show that the precoding design is equivalent to covariance matrix design of transmit signals, which can be further transformed to a convex optimization problem. To develop an algorithm to find the optimal solution, we derive the optimal structure of the covariance matrix and show that the optimal solution can be obtained via a water-filling approach. Extensive simulation results are provided to verify the performance of the proposed designs.

We investigate the information transmission in a typical communication scenario in the intelligent transportation systems (ITS), that is, infrastructures providing ITS services that intend to broadcast independent messages to the vehicles within their coverage areas. Due to the dynamic nature of wireless signal propagation links and interuser interference, it is hard to guarantee satisfactory performance by direct infrastructure to vehicle (I2V) transmissions. To solve this problem, we propose allowing certain vehicles to serve as relays to assist in the information distribution process and applying a class of finite-field network codes to efficiently use the available spectrum resources. Considering that the infrastructure information sources and all vehicles are randomly distributed following Poisson point processes, we model a general urban ITS communication network and based on the stochastic geometry we derive the probability that a vehicle can successfully recover all the desired messages from its serving infrastructure. The analytical and numerical results clearly demonstrate that our proposed network coding based relay-assisted I2V transmission can significantly improve the communication performance of the conventional direct I2V transmission strategy.

Device-to-device (D2D) communications as underlays of cellular networks facilitate diverse local services and reduce base station traffic. However, D2D communication may cause interference with the primary cellular network. To avoid this problem, the network should flexibly allocate its resources and select a proper mode for users. Here, we formulate a joint mode selection and resource allocation problem to maximize the system throughput with a minimum required rate guarantee. A mode selection and resource allocation scheme based on particle swarm optimization (PSO-MSRA) is proposed in which solutions are mapped onto particles and a fitness function embodies the constraints in a penalty function. Simulation results show its superiority over other schemes in terms of throughput and minimum required rate guarantee.

The problem of resource block allocation for downlink Long Term Evolution (LTE) systems is addressed. To maximize system throughput and fulfill quality of service (QoS) requirements among users and modulation and coding scheme (MCS) selection constraint brought by LTE, a Particle Swarm Optimization (PSO) based resource block allocation algorithm is designed. That is, we propose a coding scheme to map solutions onto particles, and find an operation to discretize the position and velocity of particles and then establish a fitness function after handling constraints through penalty function. Simulations show that the proposed algorithm outperforms in the achievement of high system throughput with QoS guarantee when resource is enough and plays a worthwhile tradeoff between QoS and throughput in overloaded systems.

This paper proposes a novel approach of cross-layer resource allocation based on Quality of Experience (QoE) for wireless video transmission. We model and formulate the cross-layer resource allocation problem in accordance with the relationship between transmission bit rate and QoE. Our objective is to ameliorate the system QoE level while guaranteeing the fairness for users. The video quality fluctuation that affects the system QoE is considered as well. The proposed algorithm designs an adaptive resource allocation approach depending on limited network resources. We use a downlink LTE system as a simulation example of a high speed downlink access system. The simulation results demonstrate that the proposed method considerably and agreeably improves system QoE more than other algorithms.

In Multi-user MIMO system, traditional user scheduling methods are usually aiming at maximizing the sum-rate capacity of the target system. Such methods select the users which have the best channel quality to service but ignore users' satisfaction. In this paper, we proposed a user satisfaction based judgment criteria, and designed a novel scheduling method in order to ensure users' satisfaction. We apply zero-forcing beam forming (ZFBF) precoding into downlink transmission, and assume perfect knowledge of CSI at the base station. Simulation results show that the algorithm achieved a higher system average user satisfaction degree while suffered a little performance loss compared with the classical SUS-ZFBF strategy.

Wireless sensor networks have been applied in many service application areas, which consist of numerous autonomous sensor nodes with limited energy. In order to optimize the wireless sensor networks performance and meanwhile expand the cellular network's service applications, mobile cellular networks and wireless sensor networks are evolving from heterogeneous to converge. However, the sensor nodes need to be informed the mobile sink's location for data collection purpose. Unfortunately, frequent location updates from multiple sinks can lead to both excessive drain of sensors' limited battery supply and increased collisions in wireless transmission. In this paper, we describe the movement direction based path selection strategy which adopts the grid structure and provides an efficient mobile sink's location updates approach. When the sinks have mobility, they only inform their current locations to the specific nodes on the previous path instead of the source nodes. We evaluate the proposed method through both analysis and extensive simulations. The results show that our proposed theory not only provides an energy-efficient location update approach, but also generates a minimal delay path for data dissemination.

MIMO, OFDMA and cooperative relaying are the key technologies in future wireless communication systems. However, under the usage of these technologies, resource allocation becomes a more crucial and challenging task. In multi-user MIMO-OFDMA relay-enhanced cellular networks, we formulate the optimal instantaneous resource allocation problem including user group selection, path selection, power allocation, and subchannel scheduling to maximize system capacity. We first propose a low-complex resource allocation algorithm named `CP-CP' under constant uniform power allocation and then use a water-filling method named `CP-AP' to allocate power among transmitting antennas. Moreover, we solve the original optimization problem efficiently by using the Jensen's inequality and propose a modified iterative water-filling algorithm named `AP-CP'. Based on `AP-CP', the `AP-AP' algorithm is proposed to allocate power adaptively not only among subchannels but also among multiple transmitting. Finally, we compare the performance of the four schemes. Our results show that allocating power among subchannels is more effective than among transmitting antennas if the average signal-to-noise radio of users is low, and vice versa. Furthermore, the `AP-AP' algorithm achieves the highest throughout especially for users near the cell edge.