Recently, unmanned aerial vehicles (UAVs) have attracted much attention due to their on-demand deployment, high mobility, and low cost. For UAVs navigating in an unknown environment, efficient environment representation is needed due to the storage limitation of the UAVs. Nonetheless, building an accurate and compact environment representation model is highly non-trivial because of the unknown shape of the obstacles and the time-consuming operations such as finding and eliminating the environmental details. To overcome these challenges, a novel vertical strip extraction algorithm is proposed to analyze the probability density function characteristics of the normalized disparity value and segment the obstacles through an adaptive size sliding window. In addition, a plane adjustment algorithm is proposed to represent the obstacle surfaces as polygonal prism profiles while minimizing the redundant obstacle information. By combining these two proposed algorithms, the depth sensor data can be converted into the multi-layer polygonal prism models in real time. Besides, a drone platform equipped with a depth sensor is developed to build the compact environment representation models in the real world. Experimental results demonstrate that the proposed scheme achieves better performance in terms of precision and storage as compared to the baseline.
As a powerful mathematical framework that allows intelligent agents to gradually learn their optimal strategies in unknown dynamic environments, reinforcement learning (RL) has found its success in many important applications. Nonetheless, a common stumbling block of RL algorithms is their low learning speed. Although different methods have been developed in literature to enhance the learning speed when special structure or prior learning experience is available, expediting RL in the general settings still remains a challenge. The Zap Q-learning is a recent breakthrough in this direction, which is shown to be an order of magnitude faster than the conventional Q-learning and its cutting-edging variants. Inspired by this exciting result, a novel algorithm, termed Glide and Zap Q-learning (G-Zap Q-learning), is proposed in this work by incorporating a novel gliding step into the learning process. The proposed algorithm is provably convergent to the optimal strategy and can further increase the learning speed of the original Zap Q-learning by several folds. In addition, it is applicable to general Markov decision processes (MDPs) and hence assumes wide applications. Simulations over both randomly generated MDPs and an exemplary application of privacy-aware task offloading in mobile-edge computing are conducted to validate the effectiveness of the proposed algorithm.
The rapid growth in scale and complexity of mobile applications fosters the development of the coded edge computing paradigm. By exploiting the redundancy in the encoded subtasks, coded edge computing enables collaborative transmission of multiple edge nodes and is promising for distributed computing in wireless fading environments. Nonetheless, to the best of our knowledge, due to challenges arising from the selection of the coding parameters, offloading strategy design for coded edge computing in general fading environments still remains open. With this consideration, the coded offloading problem is studied in this work and a delay-optimal coded offloading scheme is proposed. In particular, when the offloaded tasks are encoded by $(k,r)$ linear codes, transmission diversity gains can be obtained by performing edge node selection to mitigate fading. However, the corresponding optimization problem turns out to be a highly non-trivial non-linear mixed-integer programming. To this end, through in-depth analysis based on order statistics, it is found that the average processing delay of the offloaded tasks admits a favorable $V$ -structure with respect to the coding parameter $r$ , under arbitrary fading distribution. This key theoretic result allows us to efficiently solve the original problem using monotonic optimization. Simulations are conducted to validate our analysis and corroborate the effectiveness of the proposed scheme.
In this paper, the mobility of network nodes is explored as a new promising approach for jamming defense. To fulfill it, properly designed node motion that can intelligently adapt to the jammer's action is crucial. In our study, anti-jamming mobility control is investigated in the context of the single and multiple commodity flow problems, in the presence of intelligent mobile jammers which can respond to the evasion of legitimate nodes as well. Based on spectral graph theory, two new spectral quantities, single- and multi-weighted Cheeger constants and corresponding eigenvalue variants, are constructed to direct motions of the defender and the attacker in this dynamic adaptive competition. Both analytical and simulation results are presented to justify the effectiveness of the proposed approach. Furthermore, the proposed scheme can also be applied in cognitive radio networks to reconfigure the secondary users in the presence of mobile primary users.
An intention prediction algorithm for natural human-computer interaction based on machine vision is proposed in this paper.Firstly, the motion data of human skeletal feature point is acquired.Then, the motion data and the real-time interactive image are coupled through data processing.Meanwhile, an intention recognition model for natural human-computer interaction is built based on target feature extraction.The dominant feature weight of the operator intention is distributed by hierarchical method.A parallel scheme is adopted in this algorithm for operator intention recognition.Through experiment and data analysis, the algorithm is proved to be reliable and instrumental for improving the efficiency of natural human-computer interaction.
Driven by the exploding computing service demands from various intelligent mobile applications, an increasing amount of research efforts have been devoted to mobile edge computing (MEC). Meanwhile, unmanned aerial vehicles (UAVs) have found a great success in assisting existing wireless systems due to their flexibility and low cost. The advancements in these two closely related fields foster the development of the recently advocated UAV-assisted MEC paradigm, which is expected to bring unprecedented performance gain to the existing ground-based MEC systems. Nonetheless, existing works on UAV-assisted MEC mainly focus on the single-UAV scenarios and often assume static system states. In this paper, UAV swarm assisted MEC is considered where multiple collaborative UAVs are employed to help the terrestrial edge server to provide better edge computing services. However, the complexity of using existing methods to find the best dynamic coordination strategy in UAV swarm assisted MEC becomes intractable when the number of UAVs increases. To resolve this challenge, a novel decentralized deep reinforcement learning algorithm is proposed in this work, which can reduce the complexity by orders of magnitude. In addition, simulations are conducted to show that by using the proposed algorithm, the UAV swarm can efficiently learn a good dynamic coordination strategy and thus achieve a significantly better performance than the baseline scheme.
Communication overhead has become one of the major bottlenecks in the distributed training of modern deep neural networks. With such consideration, various quantization-based stochastic gradient descent (SGD) solvers have been proposed and widely adopted, among which signSGD with majority vote shows a promising direction because of its communication efficiency and robustness against Byzantine attackers. However, signSGD fails to converge in the presence of data heterogeneity, which is commonly observed in the emerging federated learning (FL) paradigm. In this article, a sufficient condition for the convergence of the sign-based gradient descent method is derived, based on which a novel magnitude-driven stochastic-sign-based gradient compressor is proposed to address the non-convergence issue of signSGD. The convergence of the proposed method is established in the presence of arbitrary data heterogeneity. The Byzantine resilience of sign-based gradient descent methods is quantified, and the error-feedback mechanism is further incorporated to boost the learning performance Experimental results on the MNIST dataset, the CIFAR-10 dataset, and the Tiny-ImageNet dataset corroborate the effectiveness of the proposed methods.
A fundamental assumption of link signature based security mechanisms is that the wireless signals received at two locations separated by more than half a wavelength are essentially uncorrelated. However, it has been observed that in certain circumstances (e.g., with poor scattering and/or a strong line-of-sight (LOS) component), this assumption is invalid. In this paper, a Correlation ATtack (CAT) is proposed to demonstrate the potential vulnerability of the link signature based security mechanisms in such circumstances. Based on statistical inference, CAT explicitly exploits the spatial correlations to reconstruct the legitimate link signature from the observations of multiple adversary receivers deployed in vicinity. Our findings are verified through theoretical analysis, well-known channel correlation models, and experiments on USRP platforms and GNURadio.
The recent rise of computation-intensive and large-scale intelligence applications has spurred the growth of distributed edge computing, which however is often hampered by the so-called straggling effect. To this end, coded edge computing emerges as a promising solution by creating judiciously designed redundant computations using coding theory. However, existing schemes that make the edge nodes transmit their computing results independently turn out to be sub-optimal. Although collaboration is beneficial, to the best of our knowledge, an effective cooperative transmission scheme that is universally applicable to general task encoding schemes is still missing. With this consideration, a novel cooperative transmission scheme, termed task decoding-assisted TDMA, is proposed in this work, which employs overhearing and task-decoding to enable cooperative downlink transmission in coded edge computing. Besides, an analytic bound with closed-form expression and a more accurate algorithmic bound are derived for the average downlink delay of the proposed scheme. Simulations show that the proposed scheme can reduce the downlink delay of coded edge computing by orders of magnitude.
Unmanned aerial vehicle (UAV) can act as an aerial data collector to efficiently gather fresh information in wireless sensor networks (WSNs). To support continuous data collection, the battery-powered UAV should be able to obtain energy supplements in time. In this paper, we propose a joint data gathering and energy recharging framework for age minimization in UAV-enabled WSNs. To discuss the impacts of the deployments of ground terminals and charging stations on the network age, we assume that the UAV's action follows a stochastic policy. It flies from one terminal to another with some probability when its on-board energy is sufficient, otherwise flies to the nearest charging station to recharge its battery. The UAV-aided data collection problem with energy recharging is modeled as a non-linear optimization problem. With the aid of convex programming, a stochastic optimal policy is found to minimize the network peak age. Simulation results show that the network peak age performance is greatly affected by the intensities of the ground terminals and charging stations.
'/%3e%3cuse xlink:href='%23a' transform='rotate(23.036 .093 25.536)'/%3e%3cuse xlink:href='%23a' transform='rotate(45.87 1.273 16.18)'/%3e%3cuse xlink:href='%23a' transform='rotate(69.945 .996 12.078)'/%3e%3cuse xlink:href='%23a' transform='rotate(20.66 -19.689 31.932)'/%3e%3c/svg%3e)
