Tesi etd-11112020-125044
Link copiato negli appunti
Tipo di tesi
Corso Ordinario Secondo Livello
Autore
IORI, FRANCESCO
URN
etd-11112020-125044
Titolo
Robot-Human Handover: generation of approach trajectories with Dynamic Movement Primitives
Struttura
Cl. Sc. Sperimentali - Ingegneria
Corso di studi
INGEGNERIA - INGEGNERIA
Commissione
Tutor Prof. STEFANINI, CESARE
Relatore Prof. FALOTICO, EGIDIO
Presidente Prof. FRISOLI, ANTONIO
Membro Dott. AVIZZANO, CARLO ALBERTO
Membro Prof. BUTTAZZO, GIORGIO CARLO
Membro Prof. CASTOLDI, PIERO
Membro Prof. CUCINOTTA, TOMMASO
Membro Prof. DI PASQUALE, FABRIZIO CESARE FILIPPO
Membro Prof.ssa MENCIASSI, ARIANNA
Membro Prof. MICERA, SILVESTRO
Relatore Prof. FALOTICO, EGIDIO
Presidente Prof. FRISOLI, ANTONIO
Membro Dott. AVIZZANO, CARLO ALBERTO
Membro Prof. BUTTAZZO, GIORGIO CARLO
Membro Prof. CASTOLDI, PIERO
Membro Prof. CUCINOTTA, TOMMASO
Membro Prof. DI PASQUALE, FABRIZIO CESARE FILIPPO
Membro Prof.ssa MENCIASSI, ARIANNA
Membro Prof. MICERA, SILVESTRO
Parole chiave
- Nessuna parola chiave trovata
Data inizio appello
15/12/2020;
Disponibilità
completa
Riassunto analitico
A handover is a collaborative action where an agent, the giver, passes an object to a second agent, the receiver.
This task, although seemingly effortless for humans, requires several processes: communication of intentions, coordination of the two agents, trajectory generation, grasp planning, grip force modulation, and more.
This work aims to address the generation of suitable approach trajectories, exploiting the principled framework provided by Dynamical Movement Primitives (DMP). This framework allows the inclusion of learning strategies for the generation of trajectories, by which a forcing term can be learned from a dataset of recorded human-human interactions. Finally, by adjusting the timing of the DMP the problem of coordination can be addressed naturally, without the need for a different controller.
This task, although seemingly effortless for humans, requires several processes: communication of intentions, coordination of the two agents, trajectory generation, grasp planning, grip force modulation, and more.
This work aims to address the generation of suitable approach trajectories, exploiting the principled framework provided by Dynamical Movement Primitives (DMP). This framework allows the inclusion of learning strategies for the generation of trajectories, by which a forcing term can be learned from a dataset of recorded human-human interactions. Finally, by adjusting the timing of the DMP the problem of coordination can be addressed naturally, without the need for a different controller.
File
| Nome file | Dimensione |
|---|---|
| SSSUP_TH..._Iori.pdf | 2.42 Mb |
Contatta l'autore |
|