Browsing by Author "Dario, Giacomo"

Now showing 1 - 2 of 2

Out-of-distribution- and location-aware PointNets for real-time 3D road user detection without a GPU
(2024-12) Seppänen, Alvari; Alamikkotervo, Eerik; Ojala, Risto; Dario, Giacomo; Tammi, Kari
A1 Alkuperäisartikkeli tieteellisessä aikakauslehdessä
3D road user detection is an essential task for autonomous vehicles and mobile robots, and it plays a key role, for instance, in obstacle avoidance and route planning tasks. Existing solutions for detection require expensive GPU units to run in real-time. This paper presents a light algorithm that runs in real-time without a GPU. The algorithm combines a classical point cloud proposal generator approach with a modern deep learning technique to achieve a small computational requirement and comparable accuracy to the state-of-the-art. Typical downsides of this approach, such as many out-of-distribution proposals and loss of location information, are examined, and solutions are proposed. We have evaluated the performance of the method with the KITTI dataset and with our own annotated dataset collected with a compact mobile robot platform equipped with a low-resolution LiDAR (16-channel). Our approach reaches a real-time inference on a standard CPU, unlike other solutions in the literature. Furthermore, we achieve superior speed on a GPU, which indicates that our method has a high degree of parallelism. Our method enables low-cost mobile robots to detect road users in real-time.
Towards general end-to-end sensor fusion for robot localization: implementa-tion of visual-inertial-wheel odometry
(2023-01-23) Dario, Giacomo
Perustieteiden korkeakoulu | Master's thesis
This thesis aims at generalizing a state-of-the-art end-to-end approach for robot localization in GNSS (GPS) deprived environments using a monocular camera, inertial sensors, and wheels encoder. The pipeline is trained and tested for autonomous vehicles, but the work aims to develop multimodal robot localization and observe how the method can be generalized. This thesis starts with an overview of the localization methods, structured to highlight the challenge of localization and sensor fusion, followed by a description of the state-of-the-art learning-based methods. Then, the data analysis and preprocessing are explained in the methods, as well as the structure of the pipeline, with a detailed analysis of its building blocks. Later, the results are shown and discussed, providing a comparison with the existing methods. To conclude the thesis, some observations about the presented methods and their future developments will be presented. This work has a solid industrial relevance since robustness in localization is an open problem and requires tailored engineering efforts, while having a strong research interest since it develops and expands the state of the art at the intersection between robotics and artificial intelligence.