We propose the algorithm for 3-D 6-DOF motion tracking of 3-D objects by the segment-based stereovision. Observable tracking points are selected from the sample points on the boundary segments or the nets of the curved surfaces of an object model based on the position and orientation of the object at each frame. The 3D motion parameters between two frames are obtained by searching for temporal correspondences of the tracking points and then their spatial correspondences. Repeating a constrained simple one-dimensional search for both correspondences makes it possible to track moving 3-D objects stably in real-time on a usual workstation.
This paper describes simulation results of CubeTrack Algorithm that is the spatial path generation algorithm proposed by the authors to realize a high reliability seam tracking robotic system which can operate even under poor sensing environments. CubeTrack Algorithm is evaluated on (a) Sensing error suppression characteristics and (b) Curved path restoration characteristics which are the most basic and important characteristics for path generation algorithm. Then, the relationship between N that is the number of sensing data used to generate path function by CubeTrack Algorithm and above two characteristics are obtained when sensing interval, standard deviation of sensing error and the minimum radius of curvature of the objective path are given. Based on the results, parameter design criterion appropriate to the task specification is described for CubeTrack Algorithm.
Recently, range sensors (rangefinders to sense objective profiles) have been applied widely to industrial scenes to measure the object location or shapes. In this paper, we analyse the factors to downsize the range sensor and propose one form of a very slim range sensor to realize the result of the analysis. The key component of the downsizing is a small light scanner based on the bending motion of an elastic plate with an iron leaf by electromagnetic power. We also show the calibration method which fits to the proposed range sensor. We made a prototype of the range sensor and examined its sensing capacity. As a result, the range sensor can sense profiles in the expected accuracy. The sensor is applicable not only to robotic tracking tasks, but to many measurement tasks in narrow space.
This paper discusses the manipulation for multiple objects under rolling contacts. For manipulating multiple objects, there are two key issues which do not appear in the manipulation of a single object, (1) each object motion is restricted by other objects, and (2) the contact force between objects is not controlled independently. As for (1), we consider the motion degrees of freedom of the grasped objects, and propose a necessary condition for manipulation. As for (2), we propose three indices for internal force. Finally, we show some numerical examples and an experimental results to confirm the validity of the proposed idea.
In this paper, we construct a nonlinear extended state observer (ESO) to estimate the dynamics of general multilink robot manipulators. The ESO is then applied to the high-speed/high-precision tracking control of a two-link DD robot with unknown dynamics. Control experiments are conducted which demonstrate that the proposed control method using ESO yields better tracking control performance as compared with the conventional disturbance observer (DOB) method or PD feedback.
In order to make an unmanned helicopter land on a landmark with GPS information, we have to know its position with one-meter accuracy. It is not practical to know beforehand a point with such accuracy. Therefore this paper proposes an autonomous landing system of an unmanned helicopter using GPS-based navigation globally, and also using image-based navigation locally. To search and track a landmark, we introduce an active image sensor, which can actively change the angle of view and the camera direction. The performance of the proposed system is shown by experiments using an unmanned helicopter “RMAX” for industrial use.
A grasping algorithm that takes into account both visual and tactile feedback has been developed for a system consists of a multifingered hand and a vision. The grasping process consists of two phases: the non-contact phase, i.e. the approach of the robot hand on the object, is realized using visual information while the contact phase, i.e. the touch of the object by the robot fingers is made using both visual and tactile information. To achieve these two processes we present an original algorithm that allows a multifingered hand to grasp an object using visual and tactile feedback. In this algorithm two types of motion are taken into account: the grasping motion for bringing fingers to the object surface, and the preshaping motion for changing the shape of the hand to the optimal position. Because these two types of motion are controlled based on sensory feedback, adaptiveness to changes of the environment is realized. The effectiveness of the proposed algorithm has been verified experimentally.
We have proposed a new leg-wheel hybrid mobile robot named “Roller-Walker”. Roller-Walker is a vehicle with a special foot mechanism on each leg which changes to a sole in walking mode and a passive wheel in skating mode. On rugged terrain the vehicle walks in leg mode, and on level or comparatively smooth terrain the vehicle makes wheeled locomotion by roller-skating using the passive wheels. The characteristics of Roller-Walker are: 1) it has a hybrid function but is lightweight, 2) it has the potential capability to exhibit high terrain adaptability in skating mode if the control method for roller-walking is fully investigated in the future. In this paper, the 2 leg trajectory for straight roller-walking is optimized in order to achieve the maximum constant velocity. Also a changeable ankle mechanism was integrated into the Roller-Walker system. Experiments were performed to demonstrate the validity of the concept of Roller-Walker and the results of straight roller-walk experiments are compared with those derived through simulation.
We are going to explain about sensing strategy of multiple mobile robot system, which is suitable for transport various objects whose mass and center of gravity are usually unknown. Our proposal is based on statics of manipulation by multiple fingers. We apply this analysis and sensing problem to optimal planning of multiple mobile robots. The main topic of in this paper is optimizing position of robots before the robots get started to lift an object up. Optimization in this case means maximizing probability of being successful in lifting an unknown object, without any overload situation. If lifting has resulted in failure, the robots renew the parameter of the object by their sensing data and plan their optimal lifting positions again. This optimizing process says that we can give an optimized solution according to probability of being successful, if there is causal relationship between a property of an object, a plan of robots, and a result of execution. We have verified an effect of proposed strategy by numerical computation.
Human-robot systems including interaction between human operators and robots should be designed with careful consideration for dynamic property and control ability of the human operator. In this paper, tracking tests of the human-robot system using an impedance-controlled robot are performed in order to clarify the control ability of human operators, and changes of the tracking ability and the dynamic property of the human operator are analyzed for different values of the robot impedance. From the experimental results, it is shown that the human operator tries to keep the dynamic property of the overall system as constant as possible by adjusting his or her own impedance property.
In recent years, missions for exploring small bodies such as asteroids, comets, and meteorites have received a significant attention all over the world. The rovers which hang around the surface of such small planetary bodies can provide in-situ surface observations on the various points of the target body. So the research and development of exploration rovers for small planetary bodies are very important. Under the micro gravity environment such as on the surface of small planetary bodies, traditional wheeled rovers are not expected to move effectively due to the low friction and the inevitable detachment from the surface. This paper discusses the friction-based mobility around the surface of small bodies and proposes a new mobility mechanism that drives the rover by hopping. The hop angle and the velocity are analyzed by numerical simulations, which show the effectiveness of the new mobility system.
We are developing a mobile robot, called Jijo-2, which provides office services, such as answering queries about people's location, route guidance, and delivery tasks. To smoothly interact with office people, Jijo-2 is expected to conduct natural spoken conversation. This paper describes dialogue techniques implemented on our Jijo-2 office robot, i.e. noise-free voice acquisition system by a microphone array, inference of under-specified referents and zero pronouns using the attentional states, and context-sensitive construction of semantic frames from fragmented utterances. The behavior of the dialogue system integrated with the sound source detection, navigation, and face recognition vision is demonstrated in real dialogue examples in a real office.