The maneuvering characteristics and range of motion of real aircraft are highly uncertain, which significantly increases the difficulty of trajectory prediction. To solve the problem that high-speed maneuvers and excessive trajectories in airspace cause a decrease in prediction accuracy and to find out the laws of motion hidden in a large number of real trajectories, this paper proposes a deep learning algorithm based on trajectory clustering and spatiotemporal feature extraction, which aims to better describe the regularity of aircraft movement for higher prediction accuracy. First, the abnormal trajectories in the public dataset of automatic dependent surveillance–broadcast (ADS-B) were analyzed, and to ensure the uniform sampling of trajectory data, the cleaning and interpolation of the trajectory data were performed. Then, the Hausdorff distance was used to measure the similarity between the trajectories, K-Medoids was used for clustering, and the corresponding prediction model was established according to the clustering results. Finally, a trajectory spatiotemporal feature extraction network was constructed based on a convolutional neural network (CNN) and a bidirectional long short-term memory (BiLSTM) network, and a joint attention mechanism was used to obtain the important features of the trajectory points. A large number of actual trajectory prediction experiments showed that the proposed method is more accurate than existing algorithms based on BP, LSTM, and CNN–LSTM models.

The current deep learning (DL)-based Specific Emitter Identification (SEI) methods rely heavily on the training of massive labeled data during the training process. However, the lack of labeled data in a real application would lead to a decrease in the method’s identification performance. In this paper, we propose a self-supervised method via contrast learning (SSCL), which is used to extract fingerprint features from unlabeled data. The proposed method uses large amounts of unlabeled data to constitute positive and negative pairs by designing a composition of data augmentation operations for emitter signals. Then, the pairs would be input into the neural network (NN) for feature extraction, and a contrastive loss function is introduced to drive the network to measure the similarity among data. Finally, the identification model can be completed by fixing the parameters of the feature extraction network and fine-tuning with few labeled data. The simulation experiment result shows that, after being fine-tuned, the proposed method can effectively extract fingerprint features. When the SNR is 20 dB, the identification accuracy reaches 94.45%, which is better than the current mainstream DL approaches.

Dragonfly algorithm (DA) is a recently proposed optimization algorithm based on swarm intelligence, which has been successfully applied in function optimization, feature selection, parameter adjustment, etc. However, it fails to take individual optimal position into consideration but only relies on population optimal position and 5 behaviours to update individual position, leading to low accuracy, slow convergence, and local optima. To overcome these drawbacks, Tent Chaotic Map and Population Classification Evolution Strategy-Based Dragonfly Algorithm (TPDA) is proposed. Tent chaotic map is used to initialize the population, making individuals distributed more uniformly in search space to improve population diversity and search efficiency. Population is classified according to individual fitness value, and different position update methods are adopted for different types of individuals to guide the search process and improve the ability of TPDA to jump out of local optima, thus realizing a balance between exploration and exploitation. The efficiency of TPDA has been validated by tests on 18 basic unconstrained benchmark functions. A comparative performance analysis between TPDA, Particle Swarm Optimization (PSO), DA, and Adaptive Learning Factor and Differential Evolution-Based Dragonfly Algorithm (ADDA) has been carried out. Experimental and statistical results demonstrate that TPDA gives significantly better performances compared with PSO, DA, and ADDA on the average and standard deviation in all 18 functions. The global optimization capability of TPDA on high-dimensional functions and the comparison of the time complexity of TPDA and other swarm intelligence algorithms is also verified in the paper. The results indicate that TPDA is able to perform better on optimizing functions without consuming more computational time.

A novel geolocation architecture, termed “Multiple Transponders and Multiple Receivers for Multiple Emitters Positioning System (MTRE)” is proposed in this paper. Existing Direct Position Determination (DPD) methods take advantage of a rather simple channel assumption (line of sight channels with complex path attenuations) and a simplified MUltiple SIgnal Classification (MUSIC) algorithm cost function to avoid the high dimension searching. We point out that the simplified assumption and cost function reduce the positioning accuracy because of the singularity of the array manifold in a multi-path environment. We present a DPD model for unknown signals in the presence of Multi-path Propagation (MP-DPD) in this paper. MP-DPD adds non-negative real path attenuation constraints to avoid the mistake caused by the singularity of the array manifold. The Multi-path Propagation MUSIC (MP-MUSIC) method and the Active Set Algorithm (ASA) are designed to reduce the dimension of searching. A Multi-path Propagation Maximum Likelihood (MP-ML) method is proposed in addition to overcome the limitation of MP-MUSIC in the sense of a time-sensitive application. An iterative algorithm and an approach of initial value setting are given to make the MP-ML time consumption acceptable. Numerical results validate the performances improvement of MP-MUSIC and MP-ML. A closed form of the Cramér–Rao Lower Bound (CRLB) is derived as a benchmark to evaluate the performances of MP-MUSIC and MP-ML.

A high-performance communication receiver desires a sufficient and accurate recognition of transmitted sources. In this paper, we propose a sequential decoding algorithm for the robust reception of sources with natural redundancy (NR) over the AWGN channel. To fully exploit the abundant NR in the exemplified English text sources, a causal language modeling (CLM) with a powerful Transformer decoder neural network (NN) structure is adopted for modeling the joint probability distribution. For versatility of byte-level tokenization, the UTF-8 encoding scheme is considered for the Wikipedia corpus, namely the English edition of the Wiki-40B dataset. The tree search-based M-algorithm (MA) is integrated with CLM, denoted as CLM-MA algorithm, to synthesize the a priori probability of information sequence and likelihood values of polluted symbols. Both simulation experiments and hardware platform evaluations are conducted to analyze the performance of the CLM-MA algorithm. With an adequately large M value, tremendous performance profit is achievable, especially for high-efficiency modulation levels and extremely noisy conditions. This mechanism of adopting the pre-trained NN model merely desires self-supervised learning architecture and eliminates the requirement of high-cost and accurate labels, which paves a new way for communication receiver design.

High-speed information transmission system demands more powerful communication receivers. In this paper, we integrate deep learning algorithms into constructing waveform level receiver, which is capable of harnessing natural redundancy (NR) in raw message sources. Owing to end-to-end learning paradigm of neural network (NN), the proposed NN receiver can accomplish joint symbol timing synchronization and demodulation based on oversampled waveform sequences directly. The prevalent bidirectional LSTM (BiLSTM) and Transformer structures, suitable for excavating sequential correlation of complex sources and waveform sequences, are considered as the backbone of NN receivers. Under the cases of clean label and noisy label, we conceive two novel training algorithms and interpret their principles in theory. The clean label bits are accessible from original source encoding schemes, while noisy label bits are available from the decision results of canonical receivers with independently noised waveform sequences. When noisy labels possess sufficiently small error rates, optimized NN receiver is competitive with that adopting the accurate labels. As for Markov sources and natural language text sources, simulation results show that the proposed NN receivers’ performances are superior to the classical receivers without a priori knowledge, especially when high spectral efficiency modulation schemes are encountered and random noise is significant.