It is important to estimate the number of fruits which are going to be harvested, since that information allows farmers to make key decisions regarding the production, distribution and commercialization of crops. In this project, solutions for detecting and counting apples are explored, by using detection and tracking algorithms based on neural networks. Four neural network object detection models were trained: YOLOv5, YOLOv8 and two variants of Faster R-CNN, using public as well as self-made datasets. Three tracking algorithms (StrongSort, ByteTrack and OCSort) were tested using the trained models and labeled video datasets. These datasets were developed with recordings of a real-world orchard and recordings generated with a ROS-based apple field simulator, developed as part of this project. Finally, two fitting models based on linear regression were evaluated in order to minimize prediction errors. As a conclusion of this research, it was found that the YOLOv5 and YOLOv8 performances were better than the other two, achieving a mAP50 of 0.87 and 0.85. In terms of apple tracking, the performance of OCSort stands out, reaching a HOTA metric of 60.24% and running times in the order of 30 ms.

Parkinson's Disease (PD) is a neurological disorder characterized by tremors, difficulty in movement, gait and coordination. Among other symptoms, most patients suffer from Freezing of Gait (FOG): an abrupt halt in gait, usually described by patients as “feet get glued to the ground”. Research has shown that cueing techniques help improve the patient's gait and reduce the number and duration of FOG episodes. Cueing can be defined as using external stimuli which provide temporal or spatial information to facilitate movement initiation and continuation. Although visual stimuli help alleviate FOG, state-of-the-art devices only help with scaling the gait and do not consider other known FOG triggers, such as changes in direction. Our solution (ARFoG) uses a stereo camera and state-of-the-art robotics software to dynamically build a map of the environment, localize the patient and plan trajectories to where they want to go. Once a destination is chosen via voice commands, lines indicating where to step throughout the entire trajectory are shown using Smart Glasses. With this device, the patient's focus is only on the gait movement, avoiding distractions that may trigger FoG episodes.

In recent times, unmanned autonomous sailboats have gained significant relevance due to the energy advantages they offer compared to other vessels, as their main power source is the wind, allowing them to sail for indefinite periods. One of the main challenges these sailboats face is navigating in diverse weather conditions, which has led to the need for developing an adaptive controller capable of handling these situations. This work proposes a solution to this problem by generating an adaptive controller using reinforcement learning combined with imitation learning and utilizing a simulator for training and evaluation. The conducted experiments demonstrate that the generated controllers can navigate in multiple environmental conditions and perform better than the static controller used as a reference. It is worth noting that only reinforcement learning is employed in training the resulting controllers, as imitation learning did not prove beneficial. The code can be found in the repository [1].