Diogo Miguel Matos

Abstract
With the increasing adoption of mobile robots for transporting components across several locations in industries, congestion problems appear if the movement of these robots is not correctly planned. This paper introduces a fleet management system where a central agent coordinates, plans, and supervises the fleet, mitigating the risk of deadlocks and addressing issues related to delays, deviations between the planned paths and reality, and delays in communication. The system uses the TEA* graph-based path planning algorithm to plan the paths of each agent. In conjunction with the TEA* algorithm, the concepts of supervision and graph-based environment representation are introduced. The system is based on ROS framework and allows each robot to maintain its autonomy, particularly in control and localization, while aligning its path with the plan from the central agent. The effectiveness of the proposed fleet manager is demonstrated in a real scenario where robots operate on a shop floor, showing its successful implementation.

Abstract
Autonomous Mobile Robots (AMRs) and Automated Guided Vehicles (AGVs) have emerged as key innovations in the industry world, with AMRs offering flexibility a nd adaptability for dynamic environments, while AGVs provide high accuracy for repetitive tasks; thus, this research proposes a study of fleets of both AGVs and AMRs to enhance productivity and efficiency in industrial settings. Several tests were performed where the duration of a mission, the success and collision rate, and the average number of disputes per mission were analyzed in order to obtain results. In conclusion, while AGVs tend to be more reliable and consistent in task completion, AMRs offer greater flexibility a nd speed.

Abstract
Simulators have been increasingly used on development and tests on several areas. They allow to speed up the development without damage and no extra costs. On realistic simulators, where kinematics play an important role, the modelling process should be imported for each component to be accurately simulated. Some robots are not yet modelled, as for example the Monera. This paper presents a model of a small vibration robot (Monera) that is acquired in a developed test-bed. A localisation ground truth is used to acquire the position of the Monera with actuating it. Linear and angular speeds acquired from real experiments allow to validate the proposed methodology.

Abstract
There is an increasing need for autonomous mobile robots (AMRs) in industrial environments. The capability of autonomous movement and transportation of items in industrial environments provides a significant increase in productivity and efficiency. This need, coupled with the possibility of controlling groups of heterogeneous robots, simultaneously addresses a wide range of tasks with different characteristics in the same environment, further increasing productivity and efficiency. This paper will present an implementation of a system capable of coordinating a fleet of heterogeneous robots with robustness. The implemented system must be able to plan a safe and efficient path for these different robots. To achieve this task, the TEA* (Time Enhanced A*) graph search algorithm will be used to coordinate the paths of the robots, along with a graph decomposition module that will be used to improve the efficiency and safety of this system. The project was implemented using the ROS framework and the Stage simulator. Results validate the proposed approach since the system was able to coordinate a fleet of robots in various different tests efficiently and safely, given the heterogeneity of the robots.

Abstract
Navigation is one of the most important tasks for a mobile robot and the localisation is one of its main requirements. There are several types of localisation solutions such as LiDAR, Radio-frequency and acoustic among others. The well-known line follower has been a solution used for a long time ago and still remains its application, especially in competitions for young researchers that should be captivated to the scientific and technological areas. This paper describes two methodologies to estimate the position of a robot placed on a gradient line and compares them. The Least Squares and the Machine Learning methods are used and the results applied to a real robot allow to validate the proposed approach.