Search

The 4G/5G era in the past decades has witnessed the vigorous development of Hybrid Analog and Digital Beamforming (HBF) technologies in the field of communications under cellular Multiple Input Multiple Output (MIMO) systems. As an evolution, the B5G/6G has strong visions of high-accurate positioning capabilities other than the communication quality, thus a beam alignment method towards positioning enhancement is also urgently desired in cellular systems. To this end, this paper proposed a HBF method for positioning enhancement in cellular MIMO systems. We first derive the Fisher Information for multiple-path assisted positioning as the performance criterion of positioning under a wideband channel with both precoder and combiner considered. Then a HBF strategy is proposed to optimize such criterion over multiple resources involving the transmitting power, beam and frequency dimensions, which is referred to as Sensing Beamforming. Furthermore, a Newton based heuristic method is proposed for the estimation of sensing elements (e.g. angle of arrival) from multiple paths, and the positioning results are obtained by a proposed multiple-path assisted positioning method considering the multiple path clutters in the environment. The results indicate that the proposed method can enhance the positioning performance with the accurate estimation of sensing elements.

The connected and autonomous vehicles combined with deep reinforcement learning (DRL) are capable of handling complex driving scenarios. However, due to the random exploration feature of reinforcement learning (RL), unexpected actions and collisions that would be inevitable in the real world occur during training, resulting in property damage, injury, and loss of life. To address this issue, in this article, we propose a sophisticated safe DRL in autonomous lane changing that benefits from both exploration and optimization capabilities. The key idea is first to integrate the safety constraints into the RL algorithm to limit the actions that the agent can take during training, which is implemented by designing a vehicle convex occupancy approximation to estimate the candidate action set. Then, adaptive exploration strategies are used, in which an imaginary environment based on domain randomization is built to explore areas of the action–state space where it is uncertain about the outcomes. We present a Monte Carlo tree search to replace unsafe with safe action. The twin-delayed deep deterministic policy gradient is used as the RL algorithm to train action space agents. Experimental results show that our proposed framework significantly enhances safety during the lane-change process with faster and more stable learning than the other methods.

Full frequency reuse combined with precoding is a promising solution for multibeam satellite systems (MSSs) to meet the evergrowing capacity demand. However, line-of-sight-dominant satellite-ground channels will cause severe channel correlation among the geographically clustered hotspot users (HUs), which restricts multiuser capacity over HUs. In this paper, we propose the reconfigurable intelligent surface (RIS)-aided hotspot capacity enhancement scheme for MSSs. We formulate a hotspot sum rate maximization problem with SINR constraints added on a different user set and present an alternating optimization (AO)-based algorithm for its solution. To reduce computational complexity, we propose a two-stage algorithm that sequentially optimizes RIS phase shift with manifold optimization and satellite precoding, no longer resorting to AO. The RIS phase shift design utilizes semi-orthogonal subspace maximization and pairwise channel decorrelation. This design effectively formulates the interplay between RIS phase shifts and transmit beamforming related to the SINR constraint. To circumvent high channel estimation overhead, we extend the algorithms to low-cost designs exploiting statistical channel state information. Simulation results demonstrate that our proposed RIS-aided MSS designs substantially enhance HUs’ sum rate, attributed to the RIS-enabled channel refinement mechanism. Moreover, the two-stage algorithm achieves a comparable performance to the AO-based algorithm.

The intersection network constitutes the basic skeleton of the urban traffic environment, and informative representation of the intersection plays an important role in supporting the wide variety of applications in the intelligent transportation system. In this paper, we propose the Intersection to Vector model, named Intersec2vec, to achieve an accurate, efficient, and low-dimensional representation of each intersection in the large-scale intersection network. It introduces structural and temporal modules with attention mechanisms to specifically represent the evolution of intersection features in the spatiotemporal dimension, ensuring that intersections with stronger correlations have a higher probability of co-occurrence. Furthermore, our proposed Intersec2vec model is integrated into a traffic signal control method based on deep reinforcement learning by supporting more precise sub-area divisions, named Intersec2vec-TSC. For each sub-area, Intersec2vec-TSC adopts a hierarchical structure to design agents, where the upper agent determines the common cycle length based on the overall states of the sub-area, and lower agents jointly train a centralized evaluation network to achieve optimization of green time for each intersection. We conduct the experiment on 108 signalized intersections using real online car-hailing data, and the experimental results show that our proposed method significantly improves the stability of sub-area division and reduces the waiting time of cars during peak hours compared with other comparison methods.

Vehicular crowdsensing is an efficient method of data collection in cities, benefiting from the powerful moving and sensing capabilities of intelligent vehicles. Modeling sensing capacities of vehicles and analyzing sensing effect based on sensor setup schemes are crucial in vehicular crowdsensing system. If we can infer the sensing strength of vehicles towards surrounding spatial points in advance, it would facilitate the determination of which vehicles to recruit for uploading sensing data, thereby achieving uniform and wide sensing coverage. However, the sensing strength of vehicles is not considered in the current research due to the diversity of sensors deployed on vehicles and the complexity of spatial relationships among vehicles in actual traffic scenarios. Therefore, this paper proposes a predictive vehicle recruitment method based on spatial sensing strength analysis. First, a fine-grained method is proposed to analyze the sensing strength of vehicles with different sensor setup schemes at various points in space. We fuse the sensing results of multiple vehicles and design novel metrics to evaluate the overall sensing effect, which comprehensively consider sensing strength, uniformity, and coverage ratio. The proposed method innovatively combines the perceptual properties of actual sensors and the complex spatiotemporal relationships among vehicles. Then, to guarantee the timeliness of recruiting vehicles, we distinguish moving modes to predict vehicle movements and further obtain the spatial sensing strength of vehicles at a specific moment in the future. Furthermore, combined with the sensing and communicating capabilities of the roadside infrastructure, the vehicle recruitment problem is described as an optimization problem under multiple practical constraints. To address this problem, an online heuristic algorithm is proposed based on the predicted vehicles' sensing strength. Finally, we conduct extensive simulations based on a real dataset to visualize vehicle sensing effect and verify the superiority of the proposed scheme.

The development of vehicular intelligence and networking has led to the emergence of vehicular crowdsensing as a new perception paradigm. By integrating edge computing, vehicular intelligence, and Internet of Vehicles technologies, vehicular crowdsensing is poised to have far-reaching implications in the domains of intelligent transportation, industrial sensing, and smart cities. In urban sensing scenarios, recruiting a large number of users can provide a lot of useful data, yet is costly due to expected financial incentive plans. As a solution, sparse mobile crowdsensing techniques have been proposed to collect data at a subset of sensing grids for data inference. However, the majority of these methods rely solely on explicit connections between sensing grids and do not consider implicit relations, which are crucial for accurate data inference. To achieve both high-quality data inference and cost reduction, we propose a cost-efficient vehicular crowdsensing scheme based on implicit relation-aware graph attention networks (CVC-IRGAT), which combines missing data inference with active grid selection. First, we design the IRGAT model to capture implicit and explicit relations between grids through a dual-channel mechanism of relation-aware and graph attention. Then, we design a method to assess the inferred data based on the Gaussian mixture model. Given the assessment values, a deviation information score function is proposed to measure the importance of the inferred values and model errors. Finally, we introduce active learning iterations to select the grids in accordance with this function. Extensive experiments have been conducted on real-world datasets, which demonstrate the superiority of the proposed CVC-IRGAT.