Tesi etd-10192020-101250
Link copiato negli appunti
Tipo di tesi
Dottorato
Autore
CINI, FRANCESCA
URN
etd-10192020-101250
Titolo
Human-inspired strategies to allow a fluent robot-to-human handover
Settore scientifico disciplinare
ING-IND/34
Corso di studi
Istituto di Biorobotica - BIOROBOTICS
Commissione
relatore Dott. CONTROZZI, MARCO
Parole chiave
- action coordination
- human-robot collaboration
- human-robot communication
- human-robot interaction
- object handover
- robotic grasping
- robotic manipulation
- seamless interaction
Data inizio appello
16/12/2020;
Disponibilità
completa
Riassunto analitico
As the technologies advance, robots leave their isolated workspace to enter the dynamic human populated environments. In parallel, robots are increasingly expected to collaborate in concert with humans exhibiting advanced skills of action coordination and intent communication.
One example of a daily collaborative action is the handing over of objects between a passer and a receiver. Despite being easily performed by humans, object handover is a concerted effort of prediction, perception, communication and coordination by both the agents. That’s why the implementation of an efficient and fluent human-robot handover is still an open challenge for the robotics community. Hence, my research work sought to answer the following questions: What are the main strategies used by people to coordinate their actions and intentions during an object handover task? How could robots employ these strategies?
This thesis covers different aspect of the object handover task. One of the most important one is how passers should present the object to their partner. My work showed that human passers choose and adapt their grasp type and location taking into account their partner’ needs. In particular, human passers favor precision grasps, likely to maximize dexterity and to leave enough space for the receiver to comfortably grab the object. Furthermore, human passers account for the subsequent task the receiver has to perform with the object, and they accordingly adjust their grasp location. To verify whether these strategies could provide benefits also for human-robot collaborations, I carried out a robot-to-human handover study. The latter showed that if also a robot passer grasps and presents the object taking into consideration the constraints of its partner’s following tasks, not only the collaboration results more efficient but also the human experience and perception considerably improve. These results drove me and my colleagues to open a discussion with the robotics community to emphasize the importance of task’s goal and constraints for successful robot manipulation strategies, proposing a new metric to evaluate robotic grasping, centered on the task execution.
But to efficiently handover an object, passers have to communicate to their partner the intention to interact. Similarly, robot passers must be capable to express their intentions without perturbating or brutally interrupting what the human partner is doing in that moment. But in a scenario where human agents are involved in a continuous pick and place task that could not be discontinued, when is it more appropriate to warn them that a robot requests to pass an object? My findings showed that providing a warning signal, at the beginning of a reach to grasp movement could severely interfere with the human ongoing action, increasing the number of errors made by the humans, slowing down and degrading the smoothness of their arm movement and deflecting their gaze. These disruptive interference effects strongly almost disappeared if robot warns the human partners shortly after they have picked an object, identify this as the best signaling timing. These results highlight the relevance of the robot communication timing in human-robot teamwork.
Finally, I investigated how human passer coordinates the object releasing with the receiver. My results showed that human passers use visual feedback based anticipatory control to trigger the beginning of the release, to launch the appropriate motor program, and adapt such predictions to different speeds of the receiver’s reaching out movements. When visual feedback is removed, the beginning of the passer’s release is delayed, and the passer use haptic information to modulate the dynamic of the grasp force release. With a following human-robot study I showed that robot-to-human handovers are experienced as more fluent when robot exhibits more reactive release behaviors and shorter release durations. These results suggest novel methods for programming controllers that could enable robots to hand over objects with humans in an easy, natural and efficient way.
I sincerely hope that the findings of my research can benefit the wider robotics community, with applications ranging from industrial cooperative task to household collaborative daily activities.
One example of a daily collaborative action is the handing over of objects between a passer and a receiver. Despite being easily performed by humans, object handover is a concerted effort of prediction, perception, communication and coordination by both the agents. That’s why the implementation of an efficient and fluent human-robot handover is still an open challenge for the robotics community. Hence, my research work sought to answer the following questions: What are the main strategies used by people to coordinate their actions and intentions during an object handover task? How could robots employ these strategies?
This thesis covers different aspect of the object handover task. One of the most important one is how passers should present the object to their partner. My work showed that human passers choose and adapt their grasp type and location taking into account their partner’ needs. In particular, human passers favor precision grasps, likely to maximize dexterity and to leave enough space for the receiver to comfortably grab the object. Furthermore, human passers account for the subsequent task the receiver has to perform with the object, and they accordingly adjust their grasp location. To verify whether these strategies could provide benefits also for human-robot collaborations, I carried out a robot-to-human handover study. The latter showed that if also a robot passer grasps and presents the object taking into consideration the constraints of its partner’s following tasks, not only the collaboration results more efficient but also the human experience and perception considerably improve. These results drove me and my colleagues to open a discussion with the robotics community to emphasize the importance of task’s goal and constraints for successful robot manipulation strategies, proposing a new metric to evaluate robotic grasping, centered on the task execution.
But to efficiently handover an object, passers have to communicate to their partner the intention to interact. Similarly, robot passers must be capable to express their intentions without perturbating or brutally interrupting what the human partner is doing in that moment. But in a scenario where human agents are involved in a continuous pick and place task that could not be discontinued, when is it more appropriate to warn them that a robot requests to pass an object? My findings showed that providing a warning signal, at the beginning of a reach to grasp movement could severely interfere with the human ongoing action, increasing the number of errors made by the humans, slowing down and degrading the smoothness of their arm movement and deflecting their gaze. These disruptive interference effects strongly almost disappeared if robot warns the human partners shortly after they have picked an object, identify this as the best signaling timing. These results highlight the relevance of the robot communication timing in human-robot teamwork.
Finally, I investigated how human passer coordinates the object releasing with the receiver. My results showed that human passers use visual feedback based anticipatory control to trigger the beginning of the release, to launch the appropriate motor program, and adapt such predictions to different speeds of the receiver’s reaching out movements. When visual feedback is removed, the beginning of the passer’s release is delayed, and the passer use haptic information to modulate the dynamic of the grasp force release. With a following human-robot study I showed that robot-to-human handovers are experienced as more fluent when robot exhibits more reactive release behaviors and shorter release durations. These results suggest novel methods for programming controllers that could enable robots to hand over objects with humans in an easy, natural and efficient way.
I sincerely hope that the findings of my research can benefit the wider robotics community, with applications ranging from industrial cooperative task to household collaborative daily activities.
File
| Nome file | Dimensione |
|---|---|
| Cini_Dis...on_v3.pdf | 37.40 Mb |
Contatta l'autore |
|