(→Autonomy and Perception) |
|||
| (13 versioni intermedie di uno stesso utente non sono mostrate) | |||
| Riga 2: | Riga 2: | ||
Autonomy and Perception are two key aspects of modern robotics. Our work spans both of these areas. In particular, we employ vision and radar technologies and apply them to the development of autonomous robots. Regarding autonomy, we use both classical and modern control methods, as well as reinforcement learning. | Autonomy and Perception are two key aspects of modern robotics. Our work spans both of these areas. In particular, we employ vision and radar technologies and apply them to the development of autonomous robots. Regarding autonomy, we use both classical and modern control methods, as well as reinforcement learning. | ||
| − | == | + | ==Research topics== |
===Sample Efficient RL=== | ===Sample Efficient RL=== | ||
| Riga 10: | Riga 10: | ||
Improve RL sample efficiency with two new tools: Episodic Noise and Difficulty Manager. | Improve RL sample efficiency with two new tools: Episodic Noise and Difficulty Manager. | ||
| − | [[Read more | | + | [[SampleEfficient | Read more]] |
| + | |||
| + | ===Safe Reinforcement Learning=== | ||
| + | <absHtml><iframe width="560" height="315" src="https://www.youtube.com/embed/Tpnb7Blvz3o?si=vh1qc75Im7Ufaujn" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen></iframe></absHtml> | ||
| + | |||
| + | Investigation on Safe RL algorithm performance on a realistic industrial robot (a Driveable Vertical Mast Lift). | ||
| + | |||
| + | [[SafeRL | Read more]] | ||
| + | |||
| + | ===Point2Depth: Radar Point Cloud to Depth Image=== | ||
| + | |||
| + | [[File:point2depth.png|500px]] | ||
| + | |||
| + | In this work we designed a contrastive learning-based technique to translate mmWave radar point clouds to depth images with Point2Depth. | ||
| + | |||
| + | [[Point2Depth | Read more]] | ||
| + | |||
| + | ===GT-MilliNoise: Graph transformer for point-wise denoising of indoor millimetre-wave point clouds=== | ||
| + | |||
| + | [[File:millinoise-1.gif |240px]] [[File:millinoise-2.gif|240px]] [[File:millinoise-3.gif|240px]] | ||
| + | |||
| + | GT-MilliNoise is a learning-based denoiser for indoor millimetere-wave radars point cloud. We also released MilliNoise a dataset that collects mmWave radar point cloud and labels each point with a sub-millimetric accuracy. | ||
| + | |||
| + | [[GTMillinoise | Read more]] | ||
| + | |||
| + | ===APEIRON: a Multimodal Drone Dataset Bridging Perception and Network Data=== | ||
| + | <absHtml> | ||
| + | <iframe width="560" height="315" src="https://www.youtube.com/embed/vwIJP89ua1c?si=gelot9UhHv60eYYj" title="YouTube video player" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen></iframe></iframe> | ||
| + | </absHtml> | ||
| + | |||
| + | APEIRON is a multimodal aerial dataset bridging the gap between perception and network data in outdoor environments, fostering multidisciplinary research. | ||
| + | |||
| + | [[Apeiron | Read more]] | ||
Autonomy and Perception are two key aspects of modern robotics. Our work spans both of these areas. In particular, we employ vision and radar technologies and apply them to the development of autonomous robots. Regarding autonomy, we use both classical and modern control methods, as well as reinforcement learning.
Improve RL sample efficiency with two new tools: Episodic Noise and Difficulty Manager.
Investigation on Safe RL algorithm performance on a realistic industrial robot (a Driveable Vertical Mast Lift).
In this work we designed a contrastive learning-based technique to translate mmWave radar point clouds to depth images with Point2Depth.
GT-MilliNoise is a learning-based denoiser for indoor millimetere-wave radars point cloud. We also released MilliNoise a dataset that collects mmWave radar point cloud and labels each point with a sub-millimetric accuracy.
APEIRON is a multimodal aerial dataset bridging the gap between perception and network data in outdoor environments, fostering multidisciplinary research.
Autonomy and Perception are two key aspects of modern robotics. Our work spans both of these areas. In particular, we employ vision and radar technologies and apply them to the development of autonomous robots. Regarding autonomy, we use both classical and modern control methods, as well as reinforcement learning.
Improve RL sample efficiency with two new tools: Episodic Noise and Difficulty Manager.