We are a team of competitor in ARGOS challenge. ARGOS challenge is a kind of robotics challenge that makes each team compete their robotics technologies of teleoperation or autonomous navigation and detection in gas and oil site like an offshore platform. At first in this paper, requirements of 1st competition of ARGOS challenge is introduced. Secondly, our approach and results in this event is described. And we also mention lessons learned and next development policy derived from these results in the end.
This paper proposes a motion planning method for robot hands to robustly grasp unknown shaped objects by sensor fusion of vision sensors and proximity sensors. The underlying cause of failures on vision-based grasping is the existence of prediction error especially in the blind area. In the proposed method, the information of the blind area is complemented by non-contact groping motion by robot fingers equipped with proximity sensors. The proximity sensor, which we have developed and called “Net-Structure Proximity Sensor,” realized high-speed feed-back control of the fingers to adjust the posture and distance to the target object. By integrating the proximity information and vision information, the non-contact groping motion enables to move the fingers along the object surface with keeping a constant distance, and to estimate the grasp stability before contacting the object. Motion experiments to various objects showed that the robot hand were able to approach to appropriate positions to grasp the objects independent on the shapes and surface textures. It is possible to plan approaching motions based on only visible area of the target object, and to minimize the sweeping volume of the fingers. Thus, the method is also expected to be available for bin-picking tasks in the industrial applications.
Simultaneous optimization with respect to multiple conflicting criteria is required in many fields of engineering including robotics. Such optimization in robotics often has to be achieved through experiments, which are expensive in time and/or money. Therefore, the evaluations of objective functions are scarce resources. In some cases, there are input regions on which the objectives cannot be defined, and which are unknown in advance. The existence of such unknown failure regions is a major problem in making an efficient experimental plan. This paper proposes a multiobjective optimization method which is capable of utilizing the unsuccessful samples appropriately to avoid further unnecessary experiments. By using Gaussian process classifier, the proposed method estimates the probability to fail at any inputs and use it to reduce the number of the unsuccessful evaluations. The efficacy is shown by numerical tests and a robotic experiment.
Various unmanned exploration robots such as volcano exploration robot and rescue robot have been developed. These robots require high traveling performance on rough terrain in order to move on unknown field. From the point of Rapid observation, the moving speed and cost of the operation are one of important ability of these. Therefore we developed ``Blade-type crawler'', which have high traveling performance on rough terrain keeping fast moving and simple mechanism. We conducted the field test on various rough terrains in volcanic field at Izu-Oshima. In the result, it was confirmed qualitatively that this has high traveling performance. In this research, in order to provide the design method of this mechanism, the tests in the traveling performance and climbing performance are tested with various parameters such as the blade length, wheel rotation speed and velocity of the vehicle. In addition, by blade length same as wheel diameter, high rotational wheel speed, a trend is shown that the slip rate is decreased. Hence, it is confirmed that this system is possible to perform robustness system which can travel at a constant velocity without affected road surface irregularities.
This paper discusses the resolved viscoelasticity control (RVC) method that explicitly considers the structure-variability for humanoids. In a previous report, the author proposed resolving the virtual viscoelasticity at the center of gravity into the joint viscoelasticity considering redundant degrees of freedom, and named this method as RVC. However, the author considered only the single support phase; therefore, the humanoid could be regarded as an open kinematic chain and the RVC was implemented easily. In this paper, the author extends the previous work on the RVC by considering structure-variability—the method now considers an open kinematic chain in the single support phase and a closed kinematic chain in the double support phase. This extension helps realize stable and robust walking motion on uneven terrains. The proposed method is validated using forward dynamics simulations.