Search

This research advances the field of human activity recognition (HAR) by developing a robust and interpretable deep learning model using wearable sensor data. We address seven discrete activities through a multimodal fusion architecture that synergistically combines temporal convolutional networks (TCNs), convolutional neural networks (CNNs), and long short-term memory (LSTM). Each network type caters to its strength: TCNs for temporal dependencies, CNNs for local features, and LSTMs for sequential information. A dedicated fusion layer seamlessly integrates these features, achieving a remarkable mean accuracy of 98.7% on challenging data. Finally, fivefold cross-validation is done to validate our results. We find a mean accuracy of 98.7% and a standard deviation of 0.003. In addition, we use local interpretable model-agnostic explanations (LIMs) and Shapley additive explanations (SHAP) to offer insights into the model’s decision-making process, thereby improving its transparency and fostering confidence. This study contributes by providing robust and interpretable deep learning models that can be used in various applications.

Hexapod robots are widely used in applications such as search and rescue, exploration, and surveillance due to their improved mobility and stability on uneven terrain. However, conventional hexapod robots typically suffer from issues such as heavy weight and complex design, which limits their effectiveness in certain scenarios. In this paper, we propose a new hexapod robot that addresses the issues of weight and complexity often associated with conventional hexapod robots. Our proposed design uses telescopic legs connected to pivot joints at the hips, resulting in a two-DOFs-leg configuration. The robot’s six legs are aligned with a parallelepiped body, with each leg configured with a revolute joint and a prismatic joint. With a total of twelve degrees of freedom (DOFs), the proposed robot can move in a straight line or turn. Our solution for stabilizing the robot’s horizontally oriented body on uneven terrain, using analytical kinematics, inverse kinematics, and geometric analysis, is a valuable contribution. Additionally, we present simulation of path generation and gait design for various inclined planes and stairs using a virtual prototype modeled in MATLAB/Simulink Simscape Multibody. The results of our simulated and physical prototype experiments demonstrate that the proposed design and body control algorithm effectively maintain the orientation of the robot’s body usually stable while it moves on different types of surfaces. Overall, our proposed hexapod robot using telescopic legs offers a promising solution for achieving improved mobility and stability on uneven terrain, with reduced weight and complexity compared to conventional hexapod robots.