Traditional routing techniques face considerable challenges in large-scale Low Earth Orbit (LEO) satellite networks for Internet of Things (IoT) data backhaul, such as slow convergence speed, excessive end-to-end hops, and high latency. Recent intelligent routing algorithms have achieved some progress in improving routing performance; however, significant potential remains in enhancing convergence efficiency, hop count management, and latency optimization. To overcome the insufficient adaptability of existing approaches in large-scale LEO constellation environments, this paper proposes a collaborative routing algorithm that integrates Q-learning and game theory, termed Collaborative Q-learning and Game-theoretic Routing (CoQGR). By fully exploiting satellite network topology characteristics, CoQGR deeply integrates and adapts Q-learning models with game-theoretic models. Specifically, in the Q-learning module, a greedy strategy dynamically optimizes the Q-table to accelerate convergence and establish basic routing paths. In the game-theoretic module, dynamic coalitions among satellites along the basic paths are formed based on maximum trusted communication distances, thereby enhancing the initial routes. Ultimately, under an identical maximum trusted distance constraint, unnecessary relay satellites are bypassed, reducing routing hops and further decreasing end-to-end latency. Simulation experiments are conducted based on the Starlink constellation, comparing CoQGR with recent Q-learning routing algorithms designed for large-scale constellations. The simulation results indicate that CoQGR significantly outperforms existing schemes by reducing routing convergence time, end-to-end hop counts, and average transmission latency by approximately 50%, 55.84%, and 41.39%, respectively.

Robotic graspers are essential for enhancing the efficiency and versatility of robots in grasping tasks. In this paper, we propose a novel inflatable deployable origami grasper with a rigid-flexible coupling structure. The proposed grasper can achieve multiple deployment configurations under a single pneumatic actuation, enabling both deployment and grasping operations while also allowing for passive self-folding during deflation. The design and fabrication of the grasper are presented. Then, the stiffness model for the inflatable deployable origami unit is developed based on the equivalent truss method. Experimental results show that the grasper successfully grasps objects of various shapes and sizes in both enveloping and fingertip grasping modes, using either two or four fingers. With its simple mechanical system and high deploy/fold ratio, the proposed grasper holds significant potential for applications in industrial automation and space exploration.

Mega-Low Earth Orbit (LEO) satellite constellations are characterized by highly dynamic topologies and stringent performance requirements, posing significant challenges to traditional routing strategies. Reinforcement learning (RL), particularly Q-learning, offers a promising framework for adaptive routing in such environments; however, its slow convergence limits practical deployment. To address this issue, this paper proposes an Enhanced Q-Learning Routing (EQLR) strategy that accelerates convergence through two key mechanisms: (1) pre-initialization of the Q-table leveraging the predictable orbital dynamics of satellites, thereby providing informed initial estimates; and (2) dynamic adjustment of the exploration coefficient based on Q-value variations, enabling a more efficient transition from exploration to exploitation. Simulation results demonstrate that EQLR achieves a $50 \%$ reduction in convergence time compared to conventional Q-learning, facilitating faster stabilization of routing paths under rapidly evolving network conditions. Although the average hop count and end-to-end latency remain comparable to baseline methods, EQLR offers significantly improved responsiveness, making it well-suited for large-scale, latency-sensitive satellite networks.

Vision Transformers (ViTs) are essential in computer vision but are computationally intensive, too. Model quantization, particularly to low bit-widths like 4-bit, aims to alleviate this difficulty, yet existing Post-Training Quantization (PTQ) and Quantization-Aware Training (QAT) methods exhibit significant limitations. PTQ often incurs substantial accuracy drop, while QAT achieves high accuracy but suffers from prohibitive computational costs, limited generalization to downstream tasks, training instability, and lacking of open-source codebase. To address these challenges, this paper introduces General, Practical, and Lightning Quantization (GPLQ), a novel framework designed for efficient and effective ViT quantization. GPLQ is founded on two key empirical insights: the paramount importance of activation quantization and the necessity of preserving the model's original optimization basin to maintain generalization. Consequently, GPLQ employs a sequential activation-first, weights-later strategy. Stage 1 keeps weights in FP32 while quantizing activations with a feature mimicking loss in only 1 epoch to keep it stay in the same basin, thereby preserving generalization. Stage 2 quantizes weights using a PTQ method. As a result, GPLQ is 100x faster than existing QAT methods, lowers memory footprint to levels even below FP32 training, and achieves 4-bit model performance that is highly competitive with FP32 models in terms of both accuracy on ImageNet and generalization to diverse downstream tasks, including fine-grained visual classification and object detection. We release an easy-to-use open-source toolkit supporting multiple vision tasks at [GPLQ code](https://github.com/wujx2001/GPLQ).

Large-scale low Earth orbit (LEO) satellite networks are characterized by massive node populations, rapid topology changes, and resource-constrained individual satellites. Existing routing technologies in such networks face challenges such as slow convergence, excessive end-to-end hop counts, and prolonged latency. To address these issues in applying Qlearning to large-scale LEO constellation routing, this study proposes a Maximum Trusted Distance-Based Greedy QLearning Routing (MTD-GQR) strategy. The approach first optimizes Q-tables through a greedy mechanism to accelerate convergence. It then extends beyond traditional topological constraints, where satellites link only with four adjacent neighbors, by leveraging maximum trusted distance to define inter-satellite connectivity. Simulations comparing the proposed strategy with the Dijkstra algorithm and recent Q-learning-based approaches for large-scale constellations demonstrate significant improvements: a 26.7% reduction in convergence time, a 15.8% decrease in average end-to-end hop count, and a 3.1% reduction in average latency. These results highlight the effectiveness of MTD-GQR in enhancing routing performance for large-scale LEO satellite networks.

With the recent advancements in space ad hoc networks, this paper proposes a Distributed Parity Resilient Networking Strategy (DPRNS), an improved TDMA-based approach within a spread spectrum system. To analyze directed graph connectivity in this context, we introduce the network redundancy metric, which monotonically quantifies connectivity in directed networks. Additionally, a Distributed Parity Backoff Switching Algorithm (DPBSA) is proposed to maintain a uniform distribution of parity nodes. Simulation results demonstrate that the proposed strategy significantly enhances throughput and network resilience compared to traditional TDMA and CSMA/CA approaches.

The final stage of satellite network data download is transmission from feeder satellites to Earth Stations (ESs), known as the last hop. Utilizing different scheduling strategies for this last hop can directly impact the satellite network’s throughput rate. Traditionally, researchers have focused on studying inter-satellite routing protocols or inter-satellite data offloading strategies to enhance network performance. However, these approaches do not address the issue of scheduling the last hop of large-scale Low Earth Orbit (LEO) constellations. This paper introduces a novel Last-Hop Scheduling (LHS) strategy for large-scale LEO constellations. Firstly, the LHS strategy employs a bidirectional dynamic domain to ascertain the communication domains of satellites and ESs separately. Then, the ESs interact with the satellites once in the communication domain to acquire the satellites’ information matrix and perform reconstruction. Finally, the reconstructed information matrix is subjected to the Maximum Decision Value Priority (MDVP) algorithm for the computation of scheduling commands. We simulate and validate the LHS strategy on the GW-2 constellation with 6912 satellites. The results show that LHS strategy can improve the satellite network throughput rate compared to the conventional strategy. The LHS strategy provides a new strategy for last-hop scheduling of large-scale constellations.

This paper proposes an all-optical switching space network architecture in response to the electronic bottleneck problem faced by the future demand for massive data communication in space, based on the current status of single-wavelength transmission in space laser links. The objective is to optimize the wavelength mapping of the network link, enabling the transmission and exchange of data between any nodes in the space network using a single wavelength.One of the challenges in constructing the current all-optical switching network onboard is the immaturity of all-optical wavelength conversion devices, which cannot be effectively applied in space. Additionally, the wireless transmission of single-wavelength laser links in space necessitates the differentiation of transmitting and receiving wavelengths to avoid crosstalk.In this paper, we address the immaturity of wavelength conversion devices by proposing a satellite switching network for single-wavelength transmission, focusing on satellite topology. This approach ensures efficient transmission of the entire network’s signal without the need for wavelength conversion.The feasibility of the proposed method and architecture is demonstrated, offering a viable solution for an all-optical switching space network architecture that is not constrained by wavelength conversion devices.

Successive cancellation list (SCL) decoder has high decoding performance but also require great resource consumption. In this paper, we combine the SCL decoder and the critical set to form a new decoder, which can stably reduce the number of operations by 65%$\sim$70% and maintain similar decoding performance to SCL decoder. The reduction in resource consumption depends only on the arrangement of unfrozen bit sequences, not on the quality of the channel. We also propose to combine this new decoder with adaptive algorithms. In this way, the decoding performance of the decoder can be improved beyond SCL decoders without increasing resource consumption.

Internet of Things(IoT) is developed rapidly and has permeated through many communication industries. With the increasing development of Low-Earth-Orbit(LEO) satellite industry and its advantage of border coverage, it is feasible to implement global loT with LEO satellites. Chirp Spread Spectrum(CSS), as a kind of spread spectrum technology, has never been used in LEO satellite communication systems for loT, while it has been already used in LoRa modulation. This paper is on the research of multi-terminal access protocol based on CSS modulation for LEO satellite loT. And Chirp Spread Spectrum Aloha (CSSA) as a new kind of spread spectrum aloha protocol is proposed in this paper, which can improve the system throughput effectively.

As for a new-style radar, the random pulse repetition interval (PRI) pulse waveform may be transmitted, and then the target Doppler parameter estimation precision and detection ability by using the conventional techniques may be invalid due to the irregular signal sampling. To solve these problems, this letter proposes a simple, yet effective and quasi-optimal coherent integration method with random pulse resampling. The proposed method first reconstructs the nonuniformly sampled signal into a uniform grid in the range frequency domain based on the modified sinc interpolation theory. Then, a matched filtering function related to the target chirp parameter is constructed in order to eliminate the residual coupling between range and azimuth. Finally, a moving target can be well imaged after performing the target motion compensation.

Pose-invariant facial expression recognition is quite challenging due to variations in facial appearance and self-occlusion caused by head rotations. In this paper, we propose an adversarial multi-view subspace learning method for pose-robust facial expression recognition. Specifically, a deep neural network is trained from face images of a certain pose to learn facial representations. Then, an adversarial strategy is adopted to force statistical similarity among the learned representations from facial images with different poses. Simultaneously, an expression classifier is trained on the learned pose-robust facial representations. Through adversarial learning, the proposed method leverages inherent dependencies among multiple pose facial images to construct pose-robust image representations and a classifier during training. The ensemble method is adopted to combine the predictions of multiple deep neural networks and the common expression classifier, so pose estimation is not required. Experimental results on four benchmark databases demonstrate the superiority of the proposed method to state-of-the-art works.

This letter addresses the long-time coherent integration issue for weak maneuvering target detection in a staggered pulse repetition interval (PRI) radar system. Due to the random phase fluctuation caused by the varied PRI, the complex range-azimuth coupling effects will significantly deteriorate the maneuvering target detection performance. In this letter, the nonuniformly sampled signal is efficiently reconstructed into a uniform grid based on the filterbank framework, where the uniformly sampled spectrum of a fast-moving target with Doppler ambiguity can also be reconstructed. After compensating the coupling influence caused by target motion, a moving target can be finely focused. Both simulated and real data processing results are provided to validate the effectiveness of the proposed algorithm.

Existing facial expression recognition methods either focus on pose variations or identity bias, but not both simultaneously. This paper proposes an adversarial feature learning method to address both of these issues. Specifically, the proposed method consists of five components: an encoder, an expression classifier, a pose discriminator, a subject discriminator, and a generator. An encoder extracts feature representations, and an expression classifier tries to perform facial expression recognition using the extracted feature representations. The encoder and the expression classifier are trained collaboratively, so that the extracted feature representations are discriminative for expression recognition. A pose discriminator and a subject discriminator classify the pose and the subject from the extracted feature representations respectively. They are trained adversarially with the encoder. Thus, the extracted feature representations are robust to poses and subjects. A generator reconstructs facial images to further favor the feature representations. Experiments on five benchmark databases demonstrate the superiority of the proposed method to state-of-the-art work.

Complementary codes (CCs) are a kind of two- dimensional spreading codes with both ideal auto- and cross- correlation properties. Such attractive feature enable a complementary coded code division multiple access (CC-CDMA) system outperform the traditional CDMA system with one-dimensional spreading codes in terms of interference resistance. However, it will lose such superiority over requency selective fading channels due to the strict equal gain combination requirement of CCs.This paper proves the correlation property constraints of CCs to realize an interference-free multi-user CDMA system over frequency selective fading channels. Under such constraints, a kind of CCs are onstructed and both analytical and simulation results will be presented at the end of this paper to show that such CCs provide not only interference-free feature but also full-diversity gain for a CC-CDMA system over frequency selective fading channels.

An on-chip communication network is most power efficient when it operates just below the saturation point. For any given traffic load the network can be operated in this region by adjusting frequency and voltage. For a deflective routing network we propose the design of a central controller for dynamic frequency and voltage scaling. Given history information including the load and frequency in the network, the controller adjusts the frequency and voltage such that the network operates just below the saturation point. We provide control mechanisms for continuous and discrete frequency ranges. With a discrete frequency range and taking into account voltage switching delays, we evaluate the control mechanism under stochastic, smoothly varying and very bursty traffic. Experiments demonstrate that adaptive control is very effective in minimizing power consumption at reasonable performance. Compared with a fixed high frequency network, the adaptively controlled network is significantly more power efficient. We compare it to fixed frequency networks, which are either too slow exhibiting unbounded delays, or are dimensioned for the worst case with very high frequency and are very power hungry.