Salvatore D'Avella

Mr. Salvatore D'Avella

Scuola Superiore Sant'Anna
orcid

Research Focus & Metrics

Academic Career Length

7Years (2020-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
19 February 2021 in IEEE Transactions on Automation Science and Engineering
Verified
Journal Article
Closed access
Paolo TripicchioPaolo Tripicchio
Paolo Tripicchio
MUMATTEO UNETTI
MATTEO UNETTI
SDSalvatore D'Avella
Salvatore D'Avella
Alice BuffiAlice Buffi
Alice Buffi
Andrea MotroniAndrea Motroni
Andrea Motroni
Paolo NepaPaolo Nepa
Paolo Nepa
The use of radio frequency identification (RFID) technology for the traceability of products throughout the production chain, warehouse management, and the retail network is spreading in the last years, especially in those industries in line with the concept of Industry 4.0. The last decade has seen the development of increasingly precise and high-performance methods for the localization of goods. This work proposes a reliable 2-D localization methodology that is faster and provides a competitive accuracy, concerning the state-of-the-art techniques. The proposed method leverages a phase-distance model and exploits the synthetic aperture approach and unwrapping techniques for facing phase ambiguity and multipath phenomena. Trilateration applied on consecutive phase readings allows finding hyperbolae as the localization solution space. Analytic calculus is used to compute intersections among the conics that estimate the tag position. An algorithm computes intersections quality to select the best estimation. Experimental tests are conducted to assess the quality of the proposed strategy. A mobile robot equipped with a reader antenna localizes in 2-D the tags placed in an indoor scenario and reconstructs the map of the environment through a simultaneous localization and mapping (SLAM) algorithm. Note to Practitioners —A localization technology based on passive ultrahigh-frequency (UHF) radio frequency identification (RFID) is an enabling technology for intelligent warehouses, logistics, and retails. For this reason, this work presents a novel method to estimate the tag location with high accuracy. A reader antenna is mounted on an autonomous mobile robot that can move in an indoor or outdoor environment due to a simultaneous localization and mapping (SLAM) algorithm. The motion of the antenna generates a synthetic aperture. The system receives the phase measurements from the RFID tags and generates a distance model through phase unwrapping. In such a way, the possible locations of the tags in the environment are generated, creating conics. The trilateration step is performed analytically, intersecting the obtained conics. The resulting estimations are very accurate and not computation expensive. Therefore, the proposed approach can be employed in any application where localizing objects is fundamental even when reduced computational power is available, e.g., in warehouses where the products are at known heights, or where the items are placed on a fixed infrastructure, such as high-shelves logistics, to produce the inventory of the tagged objects within each shelf.

5 October 2022 in Robotics and Computer-Integrated Manufacturing
Verified
Journal Article
Closed access
SDSalvatore D'Avella
Salvatore D'Avella
CACarlo Alberto Avizzano
Carlo Alberto Avizzano
PTPaolo Tripicchio
Paolo Tripicchio

5 January 2025 in Expert Systems with Applications
Verified
Journal Article
Closed access
XWXiaopeng Wang
Xiaopeng Wang
SDSalvatore D'Avella
Salvatore D'Avella
ZLZhimin Liang
Zhimin Liang
BZBaoxin Zhang
Baoxin Zhang
JWJuntao Wu
Juntao Wu
XYXinghua Yu
Xinghua Yu

22 February 2023 in IEEE Robotics and Automation Letters
Verified
Journal Article
Open access
SDSalvatore D'Avella
Salvatore D'Avella
ASAshok M. Sundaram
Ashok M. Sundaram
WFWerner Friedl
Werner Friedl
Paolo TripicchioPaolo Tripicchio
Paolo Tripicchio
MRMáximo A. Roa
Máximo A. Roa
Grasping a variety of objects is still an open problem in robotics, especially for cluttered scenarios. Multimodal grasping has been recognized as a promising strategy to improve the manipulation capabilities of a robotic system. This work presents a novel grasp planning algorithm for hybrid grippers that allows for multiple grasping modalities. In particular, the planner manages two-finger grasps, single or double suction grasps, and magnetic grasps. Grasps for different modalities are geometrically computed based on the cuboid and the material properties of the objects in the clutter. The presented framework is modular and can leverage any 6D pose estimation or material segmentation network as far as they satisfy the required interface. Furthermore, the planner can be applied to any (hybrid) gripper, provided the gripper clearance, finger width, and suction diameter. The approach is fast and has a low computational burden, as it uses geometric computations for grasp synthesis and selection. The performance of the system has been assessed with an experimental campaign in three manipulation scenarios of increasing difficulty using the objects of the YCB dataset and the DLR hybrid-compliant gripper.

15 October 2020 in The International Journal of Advanced Manufacturing Technology
Verified
Journal Article
Closed access
PDF file
Paolo TripicchioPaolo Tripicchio
Paolo Tripicchio
GCGerardo Camacho-Gonzalez
Gerardo Camacho-Gonzalez
SDSalvatore D'Avella
Salvatore D'Avella

27 January 2023 in IEEE Robotics and Automation Letters
Verified
Journal Article
Open access
ARAlberto Remus
Alberto Remus
SDSalvatore D'Avella
Salvatore D'Avella
FFFrancesco Di Felice
Francesco Di Felice
Paolo TripicchioPaolo Tripicchio
Paolo Tripicchio
Carlo Alberto AvizzanoCarlo Alberto Avizzano
Carlo Alberto Avizzano
Object detection and pose estimation are strict requirements for many robotic grasping and manipulation applications to endow robots with the ability to grasp objects with different properties in cluttered scenes and with various lighting conditions. This work proposes the framework i2c-net to extract the 6D pose of multiple objects belonging to different categories, starting from an instance-level pose estimation network and relying only on RGB images. The network is trained on a custom-made synthetic photo-realistic dataset, generated from some base CAD models, opportunely deformed, and enriched with real textures for domain randomization purposes. At inference time, the instance-level network is employed in combination with a 3D mesh reconstruction module, achieving category-level capabilities. Depth information is used for post-processing as a correction. Tests conducted on real objects of the YCB-V and NOCS-REAL datasets outline the high accuracy of the proposed approach.