Time-Sensitive Networking (TSN) supports mixed-criticality communication by integrating Time-Triggered (TT) and Audio Video Bridging (AVB) traffic within a unified network infrastructure. While TT flows benefit from deterministic scheduling through the Time-Aware Shaper (TAS), their presence can increase the worst-case delay (WCD) experienced by AVB traffic. However, many existing AVB-aware TT scheduling approaches incur high computational costs and lack a theoretical foundation for analyzing the impact of TT routing on AVB performance. To address these limitations, this article presents a unified routing and scheduling framework that jointly optimizes TT communication while systematically improving AVB performance. At the core of our method is a network calculus-based analysis that derives a theoretical lower bound on AVB WCD under TT interference. This bound is consistently leveraged in both the routing and scheduling stages: first, to define a flow-level metric called Impact on WCD (IoW) that guides AVB-aware routing decisions; and second, to introduce an AVB-Aware Idle Constraint that regulates TT offsets to shape residual bandwidth for AVB traffic. Extensive experiments across diverse topologies and traffic patterns demonstrate that the proposed framework significantly improves AVB schedulability and delay bounds while maintaining TT feasibility with low computational overhead. These results confirm the practicality and effectiveness of a tightly integrated approach to TSN configuration for mixed-criticality systems.

Time-Sensitive Networking (TSN) is a new technology developed from Ethernet that guarantees deterministic transmission of various types of flows, such as Time-triggered (TT) flows and Audio-video-bridging (AVB) flows, in the same network. Currently, Time-aware Shaping (TAS) and Credit-based Shaping (CBS) are widely used for scheduling hybrid flows, where the credit value in CBS is directly linked to the logical bandwidth value idleSlope , which affects the real-time performance of AVB flows. However, as the network scale increases, existing bandwidth allocation methods result in higher computation times and lower allocation success rates. In this article, the aforementioned problem is addressed by two-step: first, we design a load-balanced routing heuristic (LBRH) to improve the allocation success rate; second, we accelerate the CBS bandwidth allocation by using unified bandwidth allocation scheme, which integrates LBRH and further improves the allocation success rate. Performance evaluation in multiple test cases shows that LBRH can improve the success rate of bandwidth allocation, and the unified bandwidth allocation scheme integrating LBRH significantly reduces the overall execution time while improving the success rate of bandwidth allocation, which is more suitable for large-scale TSN networks with complex routing.

In this article, a novel adaptive self-triggered controller is constructed to handle the fault-tolerant control problem for nonlinear system with input dead zone through a funnel function method: 1) the actuator failure and the dead-zone nonlinearity are fully analyzed for the modeling purpose, we transform the system into a new model; 2) the self-triggered scheme is set up to update the control signals, which effectively economizes communication resources; 3) due to the existence of unknown control gains and unknown nonlinear functions, the adaptive neural network funnel controllers are designed, such that the tracking error is confined within the funnel boundary and all the signals of the closed-loop system are semi-globally uniformly ultimately bounded; and 4) finally, the applicability of the acquired control algorithm is demonstrated via a rigid robot manipulator system.

Despite recent advances in high-resolution (HR) face recognition, recognizing identities from low-resolution (LR) facial images remains challenging due to the absence of facial shape and detail. Current research focuses solely on reducing the distribution discrepancy between the HR and LR embeddings from the output layer, rather than thoroughly investigating the superiority of HR facial images for improved performance. In this paper, we propose a novel low-resolution face recognition method enhanced by the guidance of high-resolution facial images in both feature map space and embedding space. Specifically, in feature map space, the similarity constraint across the multilayer feature maps is adopted to align the intermediate features of facial images. Then we introduce multiple generators to recover HR images from extracted feature maps and utilize the reconstructed loss to supplement the missing facial details in LR images. In embedding space, we propose a supervised auxiliary contrastive loss to encourage the paired HR and LR embedding from the same class to be pulled together, whereas those from different classes are pushed apart. The one-to-many matching strategy and the adaptive weight adjustment strategy are applied to make the network adapt to the inputs of different resolutions. Experiments on four benchmark datasets with both synthesized and realistic LR facial images demonstrate the superiority of the proposed method to state-of-the-art.

Low-resolution face recognition is challenging due to uncertain input resolutions and the lack of distinguishing details in low-resolution (LR) facial images. Resolution-invariant representations must be learned for optimal performance. Existing methods for this task mainly minimize the distance between the representations of the low-resolution (LR) and corresponding high-resolution (HR) image pairs in a common subspace. However, these works only focus on introducing various distance metrics at the final layer and between HR-LR image pairs. They do not fully utilize the intermediate layers or multi-resolution supervision, yielding only modest performance. In this paper, we propose a novel two-stage multi-scale resolution-adaptive network to learn more robust resolution-invariant representations. In the first stage, the structural patterns and the semantic patterns are distilled from HR images to provide sufficient supervision for LR images. A curriculum learning strategy facilitates the training of HR and LR image matching, smoothly decreasing the resolution of LR images. In the second stage, a multi-resolution contrastive loss is introduced on LR images to enforce intra-class clustering and inter-class separation of the LR representations. By introducing multi-scale supervision and multi-resolution LR representation clustering, our network can produce robust representations despite uncertain input sizes. Experimental results on eight benchmark datasets demonstrate the effectiveness of the proposed method. Code will be released at https://github.com/hhwang98/TMR.