Abstract
Autonomous Mobile Robots (AMRs) have significantly transformed task management in factories, warehouses, and urban environments. These robots enhance operational efficiency, reduce labor costs, and automate various tasks. However, navigating dynamic environments with moving obstacles, such as human workers, vehicles, and machinery, remains challenging. Traditional navigation systems, which rely on static maps and predefined routes, struggle to adapt to these dynamic settings. This research addresses these limitations by developing a dynamic navigation system that improves AMR performance in industrial and urban scenarios. The system enhances the A* algorithm to account for the current positions and predicted trajectories of moving obstacles, allowing the AMR to navigate safely and efficiently. Advanced sensor technologies, such as LiDAR and stereo cameras, are utilized for real-time environmental perception. The system integrates trajectory prediction and an Artificial Potential Field (APF) method for emergency collision avoidance. The solution is implemented using the Gazebo simulator and the Robot Operating System (ROS2), ensuring real-time operation and adaptive path planning. This research aims to significantly improve AMR safety, efficiency, and adaptability in dynamic environments.

Abstract
Robotics has emerged as a cornerstone of modern society, significantly impacting diverse sectors including industry, healthcare, and defense. Among its varied applications, one of the most crucial fields is the control of rigid-structure robotic manipulators. However, conventional robotic arms are typically highly specialized and rigid in design, which limits their adaptability to different tasks and environments. One promising solution to this challenge is the development of modular robotic manipulators. This work proposes a cost-effective approach for implementing a n-Degrees-of-Freedom (DoF) manipulator. It introduces a design consisting of 3D printable links that allow for flexible assembly into custom configurations. A reconfigurable software architecture is presented, enabling automated generation of description and configuration files. This facilitates visualization, planning, and control of various custom configurations. The solution leverages the open-source Robot Operating System (ROS) as a digital twin for the modular setups. Additionally, it explores the development of hardware modules accompanying each link, facilitating independent joint control irrespective of motor type. Communication with ROS software is achieved via a CAN-based OpenCyphal network.

Abstract
Quadruped robots hold immense potential for navigating in unknown environments due to their ability to use individual footholds as well as their increased stability in uneven terrain. However, legged robots often experience limitations due to weight shifts during gait transitions. These weight shifts can cause torque peaks that exceed the capacity of the jointmotors (overdrive torque), which lead to an increased risk of mechanical failure. Through the optimization of gait parameters, it is possible to reduce these risks while maximizing performance. This paper presents the use of multi-objective optimization algorithms for gait optimization in a simulated quadruped mammal robot within the Pybullet physics engine. The main focus of the study was to compare the performance of NSGA-II, NSGA-III and U-NSGA-III in minimizing overdrive torque while maximizing travel distance. The results showed that the three algorithms solve this problem, although the NSGA-III consistently yields better results in comparison to the other versions of the NSGA algorithm.

Abstract
Several approaches have been developed over time aiming to improve the localization aspects, especially in mobile robotics. Besides the more traditional techniques, mainly based on analytical models, artificial intelligence has emerged as an interesting alternative. The current study proposes to explore the machine learning model structure optimization for pose estimation, using the RobotAtFactory 4.0 competition as the main context. Using a Bayesian Optimization-based framework, the parameters of a Multi-Layer Perceptron (MLP) model, trained to estimate the components of the 2D pose (x, y, and theta) of the robot were optimized in four different scenarios of the same context. The results obtained showed a quality improvement of up to 60% on the estimation when compared with the modes without any optimization. Another aspect observed was the different optimizations found for each model, even in the same scenario. An additional interesting result was the possibility of the reuse of optimization between scenarios, presenting an interesting approach to reduce time and computational resources.

Abstract
This paper introduces the Kabsch Marker Estimation Algorithm (KMEA), a new, robust multi-marker localization method designed for Autonomous Mobile Robots (AMRs) within Industry 4.0 (I4.0) settings. By integrating the Kabsch Algorithm, our approach significantly enhances localization robustness by aligning detected fiducial markers with their known positions. Unlike conventional methods that rely on a limited subset of visible markers, the KMEA uses all available markers, without requiring the camera's extrinsic parameters, thereby improving robustness. The algorithm was validated in an I4.0 automated warehouse mockup, with a four-stage methodology compared to a previously established marker estimation algorithm for reference. On the one hand, the results have demonstrated the KMEA's similar performance in standard controlled scenarios, with millimetric precision across a set of error metrics and a mean relative error (MRE) of less than 1%. On the other hand, KMEA, when faced with challenging test scenarios with outliers, showed significantly superior performance compared to the baseline algorithm, where it maintained a millimetric to centimetric scale in error metrics, whereas the other suffered extreme degradation. This was emphasized by the average reduced results of error metrics from 86.9% to 92% in Parts III and IV of the test methodology, respectively. These results were achieved using low-cost hardware, indicating the possibility of even greater accuracy with advanced equipment. The paper details the algorithm's development, theoretical framework, comparative advantages over existing methods, discusses the test results, and concludes with comments regarding its potential for industrial and commercial applications by its scalability and reliability.

Abstract
Accurate localization in autonomous robots enables effective decision-making within their operating environment. Various methods have been developed to address this challenge, encompassing traditional techniques, fiducial marker utilization, and machine learning approaches. This work proposes a deep-learning solution employing Convolutional Neural Networks (CNN) to tackle the localization problem, specifically in the context of the RobotAtFactory 4.0 competition. The proposed approach leverages transfer learning from the pre-trained VGG16 model to capitalize on its existing knowledge. To validate the effectiveness of the approach, a simulated scenario was employed. The experimental results demonstrated an error within the millimeter scale and rapid response times in milliseconds. Notably, the presented approach offers several advantages, including a consistent model size regardless of the number of training images utilized and the elimination of the need to know the absolute positions of the fiducial markers.