
This paper develops a novel Non-Singular Fast Terminal Sliding Mode Control (NFTSMC) scheme for two-link robotic manipulators using scalar sign function formulation. The proposed controller simultaneously addresses three critical challenges in robotic control: singularity prevention through carefully designed sliding surface exponents, complete chattering elimination via the mathematical properties of the scalar sign function, and guaranteed finite-time convergence despite dynamic uncertainties and external disturbances. The control architecture combines equivalent dynamics compensation, robust switching terms, and stabilization components, all implemented through the scalar sign function framework. Rigorous Lyapunov stability analysis establishes finite-time convergence properties, with the scalar sign function providing exact discontinuity approximation without compromising theoretical guarantees. Simulation studies on a 2-DOF robotic arm demonstrate the controller's effectiveness in achieving precise trajectory tracking while maintaining smooth control signals. The approach offers significant advantages in computational efficiency and implementation simplicity compared to conventional sliding mode controllers, making it particularly suitable for real-time robotic applications requiring both high precision and robustness.
Production, health, transport and goods systems have been severely affected by recent health crises, and continue to be so in the face of current climatic and geopolitical disruptions. In this unstable context, it is essential not only to ensure their robustness, but also to reinforce their resilience. Many works in the literature deal with the robustness of systems, but very few focus on proposals related to resilience. We therefore propose to approach this concept within the framework of a multimodal transport network modeled by a multilayer network. We develop a resilience strategy based on the curvature-core decomposition of the network and compare it with a random decision strategy and a decision strategy based on maximum transport capacities. The results show that the strategy based on the curvature-core of the multilayer network enables a better decision to be made, avoiding congestion and link saturation during reconstruction, and finally enabling a balance in the distribution of flows.
The analysis of human body movements plays a crucial role in physical rehabilitation, enabling accurate assessment and feedback for patients undergoing therapy. However, the lack of standardized datasets and benchmarks limits the development of robust AI-based evaluation models. In this paper, we propose a benchmark for human body movements in the context of physical rehabilitation exercises. A key contribution is the creation of a custom dataset comprising 400 annotated images, providing a valuable resource for evaluating motion analysis models. After extracting body landmarks, we compare multiple architectures, including a simple artificial neural network (ANN), convolutional neural networks (CNNs), recurrent neural networks (RNNs), long short-term memory (LSTM), gated recurrent units (GRU), long-term recurrent convolutional networks (LRCN), and bidirectional LSTM (BiLSTM). The experimental results demonstrate the superiority of the BiLSTM model with an accuracy equal to 91%, highlighting its potential for improving the automatic assessment of rehabilitation exercises. This benchmark serves as a foundation for future research in AI-driven rehabilitation monitoring.
Skin cancer has been recognised as a significant global health concern, with millions of cases diagnosed annually. Early detection, particularly for melanoma, has improved survival rates. In recent years, artificial intelligence (AI) techniques have gained attention in skin cancer classification and have shown potential for automated skin cancer diagnosis. However, the research mainly uses dermoscopic or macroscopic images for AI model training. This research addresses these limitations by developing a generalised deep learning model that integrates dermoscopic and macroscopic image datasets, utilising transfer learning with VGG16 to enhance accuracy and accessibility. The model is designed to generalise features from the macroscopic and dermoscopic perspective, making skin cancer diagnosis more scalable. The model was trained and tested on two diverse datasets, including HAM10000 [15] and PAD-UFES-20 [14], ensuring adaptability to clinical settings and patient populations. We achieved an average Receiver Operating Characteristic Area Under the Curve (ROC AUC) score of 0.97, which indicates a good performance of the classifier. Furthermore, we also test the model in the real-time data application by testing the model’s ability to classify the lesion from a web-camera stream for real-world scenarios.
Last-mile delivery remains one of the key challenges in modern logistics, especially given the growing demand for fast and efficient transport solutions. This study evaluates two delivery strategies: autonomous aerial drone delivery with one parcel per trip and autonomous terrestrial robot delivery with the ability to transport multiple parcels per trip. Simulation modeling is used to analyze the economic, environmental, and social aspects of these methods. The study applies Simulation of Urban Mobility software to simulate the performance of autonomous terrestrial robots in an urban environment taking into account traffic, and a mathematical model to evaluate the performance of autonomous aerial drones. Three scenarios are considered: (1) drone-only delivery, (2) robot-only delivery, and (3) mixed fleet approach combining both methods. The results show that autonomous aerial drones provide faster delivery, but autonomous terrestrial robots are a more economical and environmentally sustainable solution. The hybrid approach strikes a balance between efficiency and cost, optimizing last-mile logistics. This study contributes to the development of sustainable urban delivery models and provides practical recommendations to policy makers and logistics companies.