Advanced Robotics

Nagahama, Kotaro; Demura, Satonori; Yamazaki, Kimitoshi

doi: 10.1080/01691864.2021.1914724pmid: N/A

We propose a novel method by which a daily assistive robot can learn a new task from a user with no technical knowledge. The system is inspired by Mitate, the human capacity to communicate the movement of a tool using a different tool in a different environment. We confirmed experimentally that the system can learn a new trajectory of an object in the hand according to the online instruction by a human teacher, even if the prior knowledge about the object in the teacher's hand is not presented and the object is different from the manipulated object. These findings are not only desired in robotic support for elderly or handicapped people but also helpful for an intelligent robot system to enable human-like robust and diverse motion learning.

Kawasaki, Yosuke; Takahashi, Masaki

doi: 10.1080/01691864.2021.1919548pmid: N/A

The goal state of the robot motion involved in an action is determined by the action model based on the effect of the action and the scene state. This study focuses on serving food as a robot action; the action model of serving food is obtained based on demonstrations. The arrangement position is determined by considering the spatial relations between the object being served and the multiple selected reference objects. Therefore, it is necessary to learn the importance of the relations for the reference and the spatial relations between objects based on demonstrations. In this paper, bottom-up action modeling is proposed as an action modeling approach based on the demonstration. In this method, the spatial relations between labeled objects and the importance of the relations are learned through spatial factorization of the scene into partial relations. This approach enables the robot to determine the arrangement position intended by the demonstrator, even in unknown scenes. The effectiveness of the proposed method was evaluated by simulation, which the action of serving food.

Mori, Yoshiki; Fukuhara, Mizuki; Zhu, Mingzhu; Kinbara, Yuho; Wada, Akira; Mitsuzuka, Masahiko; Tajitsu, Yoshiro; Kawamura, Sadao

doi: 10.1080/01691864.2021.1911846pmid: N/A

In recent years, pneumatically driven soft-fingered hands constructed from polymer materials have been developed to grasp variously shaped objects. In general, the motion-control ability of these hands is limited to grasping objects. A dexterous manipulation requires pinching and in-hand manipulation. However, this is difficult because soft-fingered hands face two major problems of low positioning accuracy and vibration during high-speed motion. These problems can be solved with feedback control by using a soft sensor that does not impair finger softness. This study proposes a pneumatically driven soft-finger system with a soft-position sensor formed from photoelastic polyurethane that measures the soft-finger curvature through changes in the light intensity. A sensor feedback control for the soft finger is also established in this paper. Consequently, in a series of basic experiments, the sensor feedback control achieved precise positioning and continuous accurate path tracking. Finally, pinching motions and object orientation control were achieved by accurate continuous-path control of two soft fingers.

Magassouba, Aly; Sugiura, Komei; Nakayama, Angelica; Hirakawa, Tsubasa; Yamashita, Takayoshi; Fujiyoshi, Hironobu; Kawai, Hisashi

doi: 10.1080/01691864.2021.1913446pmid: N/A

Placing objects is a fundamental task for domestic service robots (DSRs). Thus, inferring the collision-risk before a placing motion is crucial for achieving the requested task. This problem is particularly challenging because it is necessary to predict what happens if an object is placed in a cluttered designated area. We show that a rule-based approach that uses plane detection, to detect free areas, performs poorly. To address this, we develop PonNet, which has multimodal attention branches and a self-attention mechanism to predict damaging collisions, based on RGBD images. Our method can visualize the risk of damaging collisions, which is convenient because it enables the user to understand the risk. For this purpose, we build and publish an original dataset that contains 12,000 photo-realistic images of specific placing areas, with daily life objects, in home environments. The experimental results show that our approach improves accuracy compared with the baseline methods.