In this study, we investigate the effects on personality expression by the gaze movements (eyeball, eyelid, and head movements) of a dialogue robot for three-party dialogue interactions．From a three-party multimodal dialogue data, we first analyzed the distributions of (1) gaze targets, (2) gaze durations, and (3) eyeball directions during gaze aversion．We then created gaze models based on these distributions for two of the speakers in the dataset who were found to express distinct personality．We generated the gaze motions with the models of these two speakers in the humanoid robot CommU, which has big eyes and controllable eyeball and head movements．Subjective evaluation indicated that the personalities are differently perceived for the same utterances accompanied by the motions generated by these two models．

This paper presents a planning procedure of robotic manipulators for sash insertion tasks with the perception of the frame into which the robot inserts the sash. A depth camera acquires the point cloud of the frame to estimate the position and orientation of its front flat surface. With the estimated coordinate of the frame, we apply two motion planners: one is A-star searching for the sash inserted into the frame, and another is Randomly-explored Random Trees for the manipulator that holds the sash. We demonstrate the experiments of planned inserting operation and evaluate both the precision of coordinate estimation and the success rates of insertion tasks.

There is a need for technology to investigate the large space beyond a narrow hole. An example is the exploration of the Main Stream Isolation Valve (MSIV) room at the Fukushima Daiichi Nuclear Power Plant. The authors designed and fabricated a prototype long articulated arm, LIBRA-I, for this exploration. The design requirements were that the robot should be able to pass vertically through a naarrow access hole, transport a measurement device horizontally at right angles to the hole axis, be simple, lightweight, compact, and have low power consumption. To meet these requirements, a robot configuration combining a horizontally extending arm and a 2-degree-of-freedom(DoF) joint, and a variable mass 2-DoF counterweight were adopted. As a result, LIBRA-I has a horizontal reach of 4[m], a passable hole diameter of 300[mm], a mass of 11[kg] for the arm, and 5-DoF. As a motion experiment, LIBRA-I successfully performed a horizontal extension motion from a vertically suspended condition. 3D models of the environment was successfully created by capturing video using a camera on the tip of LIBRA-I. We also measured the power consumption in static and confirmed that it was extremely low, ranging from 15.8[W] to 23.5[W].

Mobile robots that perform variable missions on behalf of humans are essential in extreme environments such as disaster sites, volcanoes, and space. On cliffs, cave surfaces, ocean floors, and asteroid surfaces, conventional wheeled robots are difficult to maintain their pose due to uneven terrain and the gravitational effect. Therefore, various grippers have been developed and installed in mobile robots to keep the robot's pose. However, because such conventional grippers are limited in the shapes they can grip, the graspable regions are discretely and randomly given in unstructured environments. In this paper, we developed a novel gripper that enables to grasp of both convex and concave shapes in terrain without pre-observation of the shape. The proposed gripping mechanism consists of a passively moving pin array with spines and actively moving holders. Furthermore, the result of gripping performance experiment on emulated terrain shapes shows that the proposed gripper can grip more various shapes than conventional grippers.

Gas pipes buried underground might be damaged by aging. Therefore, inspection is needed for safety. We have been developing a pneumatically driven robot to insert from a meter gas cock without excavation and to observe the inside conditions by locomoting in service pipes. In this paper, the practical performance of the robot was moreover enhanced, with development of a highly durable thrust module and improvement of the robustness for the bending module. By using natural rubber tubes and bellows made of polyethylene terephthalate (PET), the trust module was capable of thrusting operation for more than 5,300 times. Also, we introduced a new sealing method to prevent the air leakage from the thrust module. Furthermore, we improved the robustness for the bending module by reinforcing the connection between components and adding a cover to smooth the surface. According to a retrieval experiment from a 7[m] service line with a meter gas cock, it was confirmed that the robot could be retrieved up to 11 times (9 of which are undamaged).

In this paper, a cleaning method is designed to improve the cleaning performance of a duct cleaning mechanism using a planetary gear mechanism. The objective of this study is to develop a self-propelled robot that can clean grease accumulated in kitchen ventilation ducts to reduce the risk of duct fires. In a previous study, a planetary gear cleaning mechanism and cleaning method were designed. In an grease cleaning experiment, the robot achieved its cleaning target of an average grease thickness of 0.1 mm or less. However, the non-uniformity of the brush bristle width prevented uniform cleaning of the entire inside of the pipe, and the remaining grease could ignite, resulting in a low risk reduction effect for duct fires. In this presentation, we propose a cleaning method that can clean the inside of ducts without gaps even when spiral brushes with non-uniform brush bristle widths are used, and describe the evaluation of the cleaning performance. As a result, the cleaning rate was improved by 25.4% compared to the conventional method, and the inside of the pipe was cleaned uniformly. Therefore, it is expected that duct cleaning using this mechanism will reduce the risk of duct fires.

Service robot systems, especially household robot systems, have recently achieved adaptability in various environments and tasks by leveraging some machine learning modules. In developing and verifying such data-driven robotic systems, not only the hardware, programs, and communications but also the data and models used are components to be considered. This paper discusses effective data-driven development processes for such service robot systems by introducing and discussing case studies from the WRS2020 Partner Robot Challenge.