This paper describes an adaptive robot speed control method for safe and efficient navigation in unknown environments. Speed control is important in the following two cases. (1) When a robot moves enters a narrow free space, it needs to control the speed to avoid any collision considering the motion uncertainty. (2) When a robot enters a region whose vacancy (i.e., being free) has not been decided yet, it needs to control the speed so that it can observe the region sufficiently to be confident with the vacancy of the region. This paper proposes a simple but effective strategy for speed control that the robot selects the fastest safe speed. To adopt this strategy, we define criteria for judging a speed is safe for the above two cases. The proposed method successfully made the robot move around in unknown static environments with adaptively controlling the speed.
We discuss a conceptual design of rescue robots against nuclear-power plant accidents. We claim that the rescue robots in nuclear-power plants should have the following properties. (1) The size is small. (2) The structure is simple. (3) The number of the robots is large. This paper studies the rescue robots to rescue people in an area polluted with radioactive leakage in nuclear power institutions. In particular, we propose a rescue system which consists of a group of small mobile robots. First, small traction robots set the posture of the fainted victims to carry easily, and carry them to the safety space with the mobile robots for the stretcher composition. In this paper, we describe the produced small traction robots. And, we confirm that the robots can manipulate a 40 kg dummy doll's posture. We also examine the optimal number of robots from a perspective of working efficiency in the assumption spot.
A pattern-based approach towards the motion control of humanoid robots. which reduces its difficulty due to the strong nonlinearity, has to be assisted by an online stabilization to absorb disturbances in the real world. This paper proposes a stabilization method for humanoid robots, based on an idea of Dual Term Absorption of Disturbance, which simultaneously absorbs the error in both force pattern and kinematic pattern. These conflicting schemes are solved in accordance with the difference of time span, namely, force condition should be considered in short term, while kinematic condition tolerates to be considered rather in long term. The fact that such force-kinematic condition has to be taken into account arises from the absence of fixed points in the inertia frame. The advantage of proposed is that it allows to choose any combination of joints as modified properties, so that it is applicable for various types of robots and motions.
A new control of the whole-body motion of humanoid robots is proposed. An analogy of ZMP-COG model and carted inverted pendulum represents the core dynamics of humanoids as strong nonlinear systems. It clues to the ZMP manipulation with a large number of degrees-of-freedom in the whole-body cooperated, and to the technique of a direct handling of dynamical constraints about the reaction force from the environment. As the result, the robot body is controlled just as if it were a simple inverted pendulum. And, COG Jacobian maps the motion of such an approximate model to the real multibody system. The advantage of proposed typically appears in the robustness against large perturbations, since it is suitable for a realtime implementation with a less computational cost.
Stable grasp of an object with a multifingered hand is an important issue on manipulation. This paper addresses an analytical approach for obtaining a successfully-grasping fingertip positions on the edges of an object and its acceptable fingertip forces. The successful grasp of an object can be achieved if the resultant fingertip forces and resultant moments acting on the object are zero. At first the equilibrium condition of fingertip forces is used to select candidate combinations of the object edges touched by fingertips which may be utilizable for successfully grasping. Then, by using the moment equilibrium condition the regions of the successfully-grasping fingertip positions on the edges are obtained. It is analytically introduced that the region is bounded by boundary hyperplanes induced by the moment equilibrium condition and edge-length-limits. Two propositions are proposed for obtaining the region of the successfully-grasping finger positions. If the region cannot be found out, the candidate combination of the edges will be abandoned. Finally an algorithm is given for the fingertip forces corresponding to a fingertip position in the successfully-grasping region. Several numerical examples are conducted to verify the validity of the proposed analytical approach.
This paper is concerned with development of self-transfer-type automatic pouring robot using a cylindrical ladle. Especially, the paper focuses on the pouring control of the cylindrical ladle and sloshing suppression during ladle transfer and tilting. In order to realize fast pouring, the robot's pouring system was constructed by a feed-forward controller with a system inverse to the pouring process. In order to suppress the sloshing of the liquid in the ladle, the natural frequency of the sloshing caused by the transferring and pouring motion was identified by short-term Fourier transform. The feedback controllers in the control system were then designed by the Hybrid Shape Approach using notch filters corresponding to the identified natural frequency of sloshing. The proposed pouring and sloshing suppression controls were applied to an automatic pouring robot that had both automatic detection of the mold's status filled or unfilled with liquid and tracking control to the mold. The effectiveness of the developed control system was shown through experiments.
This paper shows that secure pinching under the gravity effect can be realized by a pair of robot fingers with hemispherical ends using a coordinated sensory-feedback signal constructed without any information of an object (e.g., center of mass, mass, length, object angle) . The proposed control signals which can be easily calculated from only encoder data of the finger angles and the finger physical parameters are adaptable to rigid objects with parallel or non-parallel flat surfaces. Numerical simulations and experimental results show that the closed-loop finger-object system finally falls into some stable state satisfying the force/torque balance, that is, into a force/torque equilibrium point. In other words, the results show that stable “blind grasping” under the situation of closing eyes and neither using any tactile sensor nor force-sensing can be actualized even by a pair of robot fingers like human pinching an object by means of the thumb and index finger, provided that preshaping of finger postures is fulfilled adequately.
Humanoid robots are high-dimensional systems; thus it is very difficult for Genetic Programming (GP) to evolve control programs for humanoid robots. In this paper, we propose a framework for GP to generate control programs for humanoid robots. The key idea in our approach is to represent target task with abstract behaviors by Genetic Programming in simplified simulation and get a prototype of the control program then interpret it with Case-Based Reasoning (CBR) in the real world environments. Accordingly, our proposed approach consists of two stages: the evolution stage and the adaptation stage. In the first stage, the prototype of the control program is evolved based on abstract behaviors in a highly simplified simulation. In the second stage, the best control program is applied to a physical robot thereby adapting it to the real world environments by using CBR. Experimental results show that our approach can generate robust control programs that can easily overcome reality gap. We declare that this approach provides a general layered framework for generating control programs for complex systems with GP.
Various applications as tele-operation, training, amusement, design supporting and other virtual reality become more immersible by force sensation. While passive force displays which use only passive elements are inherently safe, unique control strategies are required to interact with virtual objects, such as tracing over surfaces of the objects. In this paper, we introduced a basic algorithm to represent virtual walls with arbitrary orientation by passive force display. And we developed it to present polygons as triangle or circle for displaying arbitrary shape objects. Passive type systems also have a problem as it is hard to present some positions and orientations of virtual objects. Against that problem, we used a passive type force display system with redundant couple of brakes mentioned in the previous paper.
In this paper, we discuss the directions of active and passive force closures in hybrid active/passive-closure grasps. At first, we define the directions and derive the corresponding generalized force/displacement sets. Then, we show the directions of active and passive parts are orthogonal to each other. We also discuss the magnitudes of the internal forces in the manipulation of the object. In hybrid active/passive-closure grasps, there exist two kinds of magnitudes of internal forces. One is the magnitude which changes if the object moves and the geometry of the fingers changes. Such a magnitude has to be controlled. The other is the one which does not change even when the object moves. We only have to fix the joint torque component corresponding to such a magnitude. We derive the two kinds of magnitudes.
Metal spinning is a plastic forming process that forms a metal sheet by forcing the metal onto a rotating mandrel using a roller tool. Products formed by metal spinning have been inherently limited to round shapes. In this paper, we propose metal spinning of non-axisymmetric products by applying hybrid position/force control. The pushing force of the roller is regulated so that the roller can track the changing radius of the mandrel. Our forming experiment demonstrates that a thin aluminum sheet can be formed into a non-axisymmetric shape.