Search

This paper originally investigates the influence of frequency-up conversion effect on piezoelectric stack generators for high-performance energy harvesting. A compressive-mode piezoelectric generator is proposed comprised of a piezostack and a spring-mass system for electromechanical transduction and mechanical excitation, respectively. The frequency-up conversion effect is induced by amplitude truncations of the mass under external excitation. A control group, i.e. without truncation, is set for performance comparisons. Experimental results indicate that, with the frequency-up conversion, instantaneous power and average power are enhanced by over 1000 and 177 times, respectively. Each voltage response cycle consists of two stages: high-frequency high-amplitude oscillation and the rest of low-frequency low-amplitude harmonic vibration. The frequency of the former stage determines impedance matching. Due to the frequency converted up by 69 times, and intrinsic high capacitance of the stack, the matched resistance is significantly reduced from over 5 kΩ to 73.10 Ω, resulting much higher power response than that without conversion. Additionally, increase of spring stiffness causes the decrease of voltage responses and less effectiveness of frequency conversion effect. In this work we conclusively report that an instantaneous peak power output of 0.32 W can be generated from a millimeter-size piezoelectric generator, which can foster development of high-viability self-powered applications.

A combination of Fresnel law and machine learning method is proposed to identify the layer counts of 2D materials. Three indexes, which are optical contrast, red-green-blue, total color difference, are presented to illustrate and simulate the visibility of 2D materials on Si/SiO2 substrate, and the machine learning algorithms, which are k-mean clustering and k-nearest neighbors, are employed to obtain thickness database of 2D material and test the optical images of 2D materials via red-green-blue index. The results show that this method can provide fast, accurate and large-area property of 2D material. With the combination of artificial intelligence and nanoscience, this machine learning assisted method eases the workload and promotes fundamental research of 2D materials. Keywords: Two-dimensional materials, Optical contrast, Total color difference, Red-green-blue, K-means clustering, K-nearest neighbors (k-NN)

Intelligent unmanned aerial vehicle (UAV) swarm may accomplish complex tasks through cooperation, relying on inter-UAV communications. This paper aims to improve the communication performance of intelligent UAV swarm system in the presence of jamming, by multi-parameter programming and reinforcement learning. This paper considers a communication system, where the communication between a UAV swarm and the base station is jammed by multiple interferers. Compared with the existing work, the UAVs in the system can exploit degree-of-freedom in frequency, motion and antenna spatial domain to optimize the communication quality in the receiving area. This paper proposes a modified Q-Learning algorithm based on multi-parameter programming, where a cost is introduced to strike a balance between the motion and communication performance of the UAVs. The simulation results show the effectiveness of the algorithm.

With the aid of software-defined radio (SDR) technologies, the recent progress in automatic modulation classification of high accuracy using deep neural networks (DNNs) may play an increasingly important role in spectrum monitoring. However, the computational demand brought by the stochastic gradient descent method in the training process is unbearable in edge devices. In this paper, we proposed a lightweight trained neural network for an Internet of thing (IoT) chip by separating the learning process from the inference process. Besides, we analyzed the influence of the size of the convolution kernel and the length of the sampling samples on the calculation and memory usage of the neural network Experimental results on actual datasets show that the proposed network occupies about 700kb of memory, and requires a single sample floating-point operation per second (FLOPs) of about 400k, while achieving high accuracy of 95.58% at a signal-to-noise ratio (SNR) from -10 to 19dB.

The recent progress in robotic communications and networking demands novel simulation platforms for development and verification, especially for deep-coupled control and communication co-design in networked robotic systems (NRS). This paper analyzes the challenges and requirements of robotic communications and networking simulations. It is shown that the popular tools of software simulation and field test may hardly fulfil the fidelity, reproductivity and flexibility demands. Therefore, this paper proposed a novel hardware-in-the-loop (HIL) architecture for networked robotic system simulation platform (NRS-SP), which may strike a beneficial tradeoff between fidelity, reproductivity and flexibility. The enabling technologies of NRS-SP are discussed and a full-fledged prototype is implemented to illustrates its potential to support the simulation, development and verification of next generation networked robotic systems.

In this paper, we originally propose an array of piezoelectric generators harnessing energy from suspension systems of oceanic vehicles based on piezoelectric stacks. The design mainly includes base and load platforms, six single vibration suppression units which contain piezoelectric units, flexible joints, springs, and force amplifiers. The piezoelectric stacks harvest energy dissipated into the environment from vibrations in the suspension systems with the merits of low internal resistance and high capacitance so that we can achieve high power output. Experimental results show under the excitation of 0.4 g and a workload of 0.5 kg at resonance, six generators can generate a total peak power of over 0.57 W. The generators can light up an array of over 200 LED diodes. The generators prove capable of charging millifarad or farad-scale capacitors to saturation in 10 s or less. Upon comparison, the in-series connection is more recommendable as much more charges can be stored in the capacitors than that of the parallel connection. We also perform multiple experiments on oceanic vehicles at sea at different speed, and it can fully charge 100 mF capacitors in 80 s. This research can be of great significance for harvesting vibration energy from the suspension system of sea vehicles.

In this paper, a cognitive jammer is developed which can adaptively and optimally jam the radar and protect the target from being detected. The interaction of the cognitive jammer and the environment is modeled as a finite Markov Decision Process based on the framework of reinforcement learning. Q-learning algorithm is used to solve the optimal jamming frequency selection problem. After several interactions, the jammer can learn the radar's strategy and optimize its jamming frequency to achieve a larger jamming-plus-noise to signal ratio (JNSR) during the whole process. Numerical results are given to illustrate the effectiveness of the proposed method. Compared with a random jamming frequency selection method, the JNSR of the proposed method is significantly larger.

Frequency hopping communication and spread spectrum communication are the common means of anti-jamming, and these means have their own limitations. In the case of unpredictable jamming, reinforcement learning algorithm can provide good guidance for communication frequency. Firstly, this paper introduces the framework of communication system and the jamming modes commonly used in communication. Then the basic principle of Q-Learning algorithm is briefly introduced. Meanwhile, the meanings of the parameters in the algorithm are explained. In the part of numerical simulation, different jamming modes are used to test the performance of the algorithm. It proves that Q-Learning algorithm has good memory in sweeping, sinusoidal and Gaussian random jamming. It can provide better policy for communication frequency in real time.

Cognitive radio (CR) technology, which enables unlicensed secondary users to opportunistically access the unused licensed spectrum, has attracted more and more attention from both academia and industry due to its potential to significantly improve the spectrum utilization. Considering both temporal and spatial variations of spectrum availability, this paper focuses on improving the energy efficiency in CR networks by opportunistically serving the delay-tolerant data only when enough spectrum is available. Based on this idea, a stochastic optimization problem is formulated to integrate the power control, link scheduling, and routing, which minimizes the expected power consumption while guaranteeing the system stability. To obtain the solution, we use the Lyapunov optimization technique and design an online algorithm, which solves a sub-problem without future knowledge of the related stochastic models (e.g., random data arrival and spectrum supply). Besides, in view of the NP-hardness of the sub-problem, we also develop a heuristic algorithm based on branch-and-bound framework to obtain the approximate solution with low computing complexity. Theoretical analysis shows that our algorithm offers an explicit tradeoff between energy consumption and delay performance. Numerical results also confirm the effectiveness of our solutions.

In order to expand the scale of wireless multi-hop cooperative network, improve the network fault tolerance and reduce network establishment delay, we combine the centralized and distributed scheduling mechanisms. When the network scale is small, the distributed mechanism can reduce the scheduling delay. When the network scale is large, the centralized mechanism can improve network stability. The hybrid scheduling combines the advantages of distributed and centralized scheduling mechanisms. We fully consider network establishment delay, network access delay, the three-way handshake delay of distributed scheduling, and the route establishment delay of centralized scheduling mechanism and propose the slot allocation scheme of hybrid scheduling mechanism. Simulation results verify the effectiveness of the proposed scheme.

Distributed multi-hop cooperative network is a new network architecture that can solve the last mile bottleneck problem, with larger network coverage, higher reliability, higher utilization, self-configuring and self-healing. The competition mechanism is one of the key mechanisms that affects the performance of Distributed multi-hop cooperative networks. This paper focuses on the research of the Distributed multi-hop cooperative network competing mechanism. Regarding the fairness problem in existing Mesh Election algorithms, this paper puts forward an improved Mesh Election algorithm. Simulation is conducted on based on NS2 platform to verify the effectiveness of the proposed algorithm.

With the increasing demand for communication quality, improving the service quality has become the focus and difficulty of the research. The traditional method has its limitations. The cross layer design method can break the barriers between layers, and ensure the quality of service transmission. This article proposes a cross-layer scheme called Ad hoc On-demand Distance Vector Routing Cross-layer Scheme(AODV-CRS) for network layer AODV routing protocol and MAC layer distributed scheduling protocol, which uses the link resource in the process of selecting the best route. In addition, the content used to apply and allocate bandwidth will be put into the routing message. It will accomplish resource allocation while establishing the route. The delivery ratio, time delay and overhead are analyzed using the NS2 simulation platform. Simulation results verify the effectiveness and reliability of AODV-CRS.

The networked unmanned aerial vehicle (UAV) radar system may exploit inter-UAV cooperation for enhancing information acquisition capabilities, meanwhile its inter-UAV communications may be interfered with by external jammers. This paper is devoted to quantifying and optimizing the anti-jamming performance of networked UAV radar systems in adversarial electromagnetic environments. Firstly, instead of using the conventional metric of signal-to-interference ratio (SIR), this paper explores use of the theory of radar information representation as the basis of evaluating the information acquisition capabilities of the networked UAV radar systems. Secondly, this paper proposes a modified Q-Learning method based on double greedy algorithm to optimize the anti-jamming performance of the networked UAV radar systems, through joint programming in the frequency-motion-antenna domain. Simulation results prove the effectiveness of the algorithm in two different networking scenarios.