This work presents a study of the frictional effect on the stability of a three-rigid link object manipulated by two cooperative robot arms in a plane. It is supposed that the three-rigid link object interacts with the two robot arms at multi contact points to perform the nonprehensile manipulation of the object. The object is figured out in a way such that one of the arms is in contact with two links of the object, while the other arm is free to slide along the third link. The effect of changing frictional forces at the contact points as well as their directions on the stability of the manipulated object is to be explored by defining the system constraints. A definition of what is called the “Stability Margin” is obtained. For the same orientation of the object links, the Stability Margin is a region of equilibrium contact points between the sliding arm and the corresponding object link where the robot arm could be placed in without dropping the object. Later, an experimental system consisting of a three-rigid link object and a two planar robot arms manipulation system is equipped to verify the concept.

In aging societies, there is a strong demand for robotics to tackle problems caused by the aging population. Patient transfer, such as lifting and moving a bedridden patient from a bed to a wheelchair and back, is one of the most physically challenging tasks in nursing care, the burden of which should be reduced by the introduction of robot technologies. We have developed a new prototype robot named RIBA with human-type arms that is designed to perform heavy physical tasks requiring human contact, and we succeeded in transferring a human from a bed to a wheelchair and back. To use RIBA in changeable and realistic environments, cooperation between the caregiver and the robot is required. The caregiver takes responsibility for monitoring the environment and determining suitable actions, while the robot undertakes hard physical tasks. The instructions can be intuitively given by the caregiver to RIBA through tactile sensors using a newly proposed method named tactile guidance. In the present paper, we describe RIBA's design concept, its basic specifications, and the tactile guidance method. Experiments including the transfer of humans are also reported.

To improve the flexibility of robotic learning, it is important to realize an ability to generate a hierarchical structure. This paper proposes a learning framework which can dynamically change the planning space depending on the structure of tasks. Synchronous motion information is utilized to generate modes and different modes correspond to different hierarchical structure of the controller. This enables efficient task planning and control using low-dimensional space. An object manipulation task is tested as an application, where an object is found and used as a tool (or as a part of the body) to extend the ability of the robot. The proposed framework is expected to be a basic learning model to account for body image acquisition including tool affordances.

The paper deals with the motion planning for a rolling system with limited contact area. The system under consideration is represented by a hemispherical object that can roll without slipping or spinning on the plane. Under the constraints imposed on the size of the contact area, the construction of motion can be regarded as a problem of parallel parking in a finite number of movement steps. A motion planning algorithm, realizing the movement steps by tracing smooth figure eights on the hemisphere, is introduced. To generate asymmetric figure eights, a generalization of the Viviani curve is proposed. An exceptional case of the algorithm, corresponding to a spinning maneuver, is constructed with the use of the Cassini curve. The convergence of the algorithm is analyzed and its computational feasibility is verified under simulation.

This paper presents a learning approach to acquisition of manipulation control for carrying an object accompanied with stick/slip mode switchings. The proposed learning architecture has hierarchical structure. The upper layer determines a global trajectory and a sequence of mode switchings by reinforcement learning while giving the lower layer commands about small motions keeping (sliding or sticking) modes. The lower layer realizes a desired small regional motion using the estimated boundary information about mode switches by support vector machine (SVM). Mode transition data are collected off-line for SVM learning. The proposed learning framework is applied to an object moving task with two-DOF manipulator, where the desired configuration of the object is specified so that stick/slip mode switchings are inevitable to accomplish the given task. Simulation showed that the proposed learning architecture realized an object control with stick/slip mode switching.

The paper deals with modeling of human-like reaching movements using optimal control theory. Typically, in the construction of optimization models, capturing the invariant features of human movements, the main emphasis is placed on the form of the optimality criterion. However, the boundary conditions are also an important part of the optimization problem. Considering reaching movements in the manipulation of flexible objects, we first show that the conventional imposition of the boundary conditions does not always produce a good match to the experimental data featuring the acceleration jumps in highly dynamic tasks. To explain the acceleration jumps, we reformulate the problem, using the concept of natural boundary conditions, and show that it improves the prediction of the experimental data. Finally, it is suggested how not only the acceleration but all the boundary conditions can placed in a natural way.

The paper deals with the motion planning problem for a rolling sphere with limited contact area. The system under consideration is represented by a hemispherical object that can roll without slipping or spinning on the plane. Under the constraints imposed on the size of the contact area, the construction of motion can be regarded as a problem of parallel parking in a finite number of movement steps. A motion strategy, realizing the movement steps by tracing generalized figure eights on the hemisphere, is introduced. Two different algorithms for this motion strategy, the circle-based and the generalized Viviani-curve-based ones, are proposed. The convergence of the algorithms is analyzed, and the computational feasibility of these algorithms is verified under simulation.

The paper deals with motion planning for rolling-based systems with limited contact area. In a simplified formulation, the driving principle for such systems is based on controlling the position of the center of mass of the object, exploiting non-holonomic rolling constraint to propel the hemisphere. Two geometric algorithms for motion planning are formulated, analyzed, and tested under simulation. The simulation results show the possibility of moving the system to the desired configurations by steering the contact point on the hemisphere by generalized figure eights represented by circles and Viviani's curves.

The paper deals with the dynamics and motion planning for a rolling system with variable inertia. In a simplified formulation, the driving principle for the system under consideration is based on controlling the position of the center of mass of the object, exploiting non-holonomic rolling constraint to propel the hemisphere. We derive the dynamic model of this system and establish the condition of motion realizability. Providing that this condition holds true, we derive an approximate analytical solution for the trajectory of the particle, driving the system along a pre-specified path in the contact coordinates. The approximation is based on ignoring the quadratic velocity terms in the dynamic equations.

The paper deals with motion planning for rolling-based systems with limited contact area. To understand the problem we resort to a simplified quasi-static model in which the locomotion object is represented by a mass point. In this formulation, the driving principle is based on controlling the position of the center of mass of the object, exploiting nonholonomic rolling constraint to propel the hemisphere. Two motion planning algorithms are proposed. The algorithms are tested under simulation. The simulation results show the possibility of steering the locomotion system to the desired configurations by moving the center of mass through multiple generalized figure eights on the main hemisphere plane

Our goal is to create advanced engineering systems such as a soft human interactive robot. The robot developed here is named RI-MAN. RI-MAN exhibits the skill and ability to realize human care and welfare tasks. RI-MAN can search out a specific person in real time by fuing audio and visual information, and understand human speech based on a sound recognition function. In addition, RI-MAN's body is coverd with soft touch sensors, and RI-MAN can react to the amplitude and location of external forces. Using all these sensor functions, RI-MAN can successfully follow human commands and hold up a dummy of the same size as an adult human. RI-MAN will become an invaluable partner robot.

The paper presents an analysis of human reaching movements in the manipulation of flexible objects. Two models, the minimum hand jerk and the minimum driving force-change, are derived and their basic features are analyzed. It is shown that the first model features two-phased hand velocity profiles, while in the second models there are multiple phases. The analysis of the phase transitions for the models considered is done in the analytical form. The results of this analysis can be helpful in the design of experimental scenarios for the verification of the theoretical models. Finally, we present some initial experimental results and analyze the applicability of the models developed in this paper

Human being can decide his walking to auto-adapt to environment changes. Comparing with this, the autonomy of biped robots is inefficient or remains unsolved. One basic reason for this is the less of a unified modeling framework for various types of biped motion in varying environment which allows on-line deriving of the control input In this paper, we propose to recast types of biped motion in the framework of hybrid system, express it as a mixed logic dynamical model. Such a point of view possesses the advantage that it encompasses all the motions (such as single support motion and impact phenomenon) into a unique one, thus allow the control system to be design optimally even though the impact occurs as the interior point of the optimization horizon

The paper presents an analysis of human reaching movements in manipulation of flexible objects. To predict the trajectory of human hand, a minimum crackle criterion has been recently introduced in literature. A different approach is explored in this paper. To explain the trajectory formation, we resort to the minimum hand jerk criterion. First, we show that this criterion matches well experimental data available in literature. Next, we argue that, contrary to the minimum crackle criterion, the minimum hand jerk criterion produces bounded hand velocity profiles for multimass flexible objects. Finally, we present initial experimental results confirming the applicability of the minimum hand jerk criterion to the prediction of reaching movements with multimass objects.

The paper presents an analysis of human reaching movements in manipulation of flexible objects. To predict the trajectory of human hand, a minimum crackle criterion has been recently proposed in literature. A different approach is explored in this paper. To explain the trajectory formation, we resort to the minimum hand jerk criterion. First, we show that this criterion matches well experimental data available in literature. Next, we argue that, contrary to the minimum crackle criterion, the minimum hand jerk criterion produces bounded hand velocity profiles for multi-mass flexible objects. Finally, we present initial experimental results confirming the applicability of the minimum hand jerk criterion in manipulation of multi-mass objects.

Constrained human movements are considered in this paper. The external constraints decrease the mobility of the human arm and lead to the redundancy in the distribution of the interaction force between the arm joints. To investigate the trajectory formation in the constrained human movements, we first develop a novel experimental system with interchangeable geometric constraints. Then, we examine the trajectory of human arm for an elliptic constraint. To clarify the trajectory formation in constrained point-to-point motions, we analyze experimental data and test them against predictions obtained by conventional criteria of optimality. It is found in the comparative analysis that the best prediction is given by the minimum muscle force change criterion.

In this paper, by taking a multi-contact planar manipulation system as an example, we propose a modeling and control approach based on hybrid system theory, whereby a dexterous manipulation task is formulated as a mixed logic dynamical (MLD) model. A model predictive control (MPC) is used to handling the control objectives of MLD system, which can be solved using powerful mixed integer quadric programming (MIQP) algorithm. The validation of the proposed approach is shown by simulation.

This paper presents a feature extraction method for robotic motion learning that optimizes image resolution to the task, thereby minimizing computation time. It utilizes mean-shift algorithms and principal component analysis for feature extraction, reinforcement learning for motion learning, and trial and error for finding the appropriate resolution. When applied to a manipulator pushing an object, the resolution adjustment method reduces the task time from one minute to 21 seconds.

This paper proposes a reinforcement learning method for dynamic control problems with holonomic constraints. The learning method is applicable to problems where the actual motion of the system is restricted to lower-dimensional submanifolds, so long as certain conditions are satisfied. Such dynamic control problems occur in robotic manipulation, which usually includes some holonomic constraints between the object and the robot or the environment. By introducing nonlinear mapping to one-dimensional space and approximating the boundary of a discontinuous reward function, the proposed method results in effective learning. The method is evaluated in a one degree of freedom object rotating task with contact force considerations. The effectiveness of the proposed learning method was verified by comparison to ordinal Q-learning and Dyna without the proposed mapping method.

It is known that the motor control system of a human is a natural hybrid system. A human uses his predictive function based on an internal model together with feedback junction for motion tracking, which is considered as a motor control model of a cerebellum. However, in control engineering field, the understanding of hybrid systems control is rather limited at present. The system theoretic studies such as analysis of controllability and trackability are very few in literature. In this study, we consider the tracking control problem of hybrid plants represented by MLD model to follow a family of reference signals produced by an external generator. New algorithms are presented. The internal model principle of continuous system is extended to hybrid systems so as to solve the problem.

This paper studies modeling and bio-mimetic control of a 3D 8-dof whole arm cooperative manipulation system using sensitive skin. A sphere is considered here as a manipulated object. The control law is designed based on integration of voluntary and reflex movements considering the system's redundancy. The voluntary task for holding the object is realized by impedance control at the four contact points with the object using the contact force information from the sensitive skin. The reflection on the other hand is introduced as a regulation problem of the direction between the points of end-effectors and elbows. The solution for the redundant control is formulated and derived from the optimization point of view. The validity of the proposed method is investigated by numerical simulations.

Hierarchical structures of multiple modules are investigated to solve large and complex problems. Many studies have effectively utilized the abstraction of successive states, but there have been no sufficiently detailed discussions of the overlap between these abstracted states and continuity between multiple modules. Here, we propose organization of the evaluation space considering the continuity of multiple rewards. Multiple reward functions are applied to the planning scheme and each reward function provides different discretization of the state space. Based on the use of multiple resolutions, we propose integration of the evaluation space. The objective of this integration was to reduce the number of discrete state sets used to calculate the value function of the task. We report here an example of effective use of multiple resolutions based on multiple rewards is effective for planning in a two-DOF manipulator motion planning problem with redundant tactile sensors.

A distributed PHANToM-based system for collaborative haptic visualization of a VR crank is presented. Physical IDOF crank model providing realistic kinaesthetic sensations of inertia and viscosity is described. Theoretical and experimental kinematic trajectory patterns are compared for the case of cooperative two-arm human movements. Nonvisual visualization applications of the system are discussed.

Opening a door, turning a steering wheel, rotating a coffee mill are typical examples of human movements constrained by the external environment. The constraints decrease the mobility of the human arm and leads to the redundancy in the distribution of the interaction force between the arm joints. Due to the redundancy of the force actuation in the constrained motions, there is infinite number of ways to form the trajectory of the arm. However, human forms the hand trajectory in a unique way. How does human resolve the redundancy of the constrained motions and specify the hand trajectory? To investigate these problems, we examine the trajectory of human arm in a crank rotation task. To explain the trajectory formation in constrained point-to-point motions, we formulate an optimal control problem and propose a novel criterion minimizing the hand contact force change and muscle force change over the time of movement. The simulation results are compared with human motion and force profiles obtained experimentally. It is shown that the novel criterion captures the characteristics of the human constrained motion much more satisfactory than conventional criteria accepted in the research community.

Unlike the present manipulator control technologies that operate objects only by the robots' end-effectors, human beings can perform whole body manipulation flexibly and easily. Bio-mimetic research of such skillful human motor behavior is important not only for deeper understanding of human sensations and nervous control functions but also for developing of higher level robots. In this paper, we first formulate the basic model of multipoints whole body interaction between a robot manipulator and its object. We then study three control approaches by considering the limitations of force feedback, the complexity of control algorithms. As an example, we consider a 2-link planar manipulator that operates a circle object with dynamic arm friction. We show a computer simulation algorithm, and compare the results of each control methods, numerically.

The paper deals with the problem of the trajectory formation in human movements constrained by the external environment. To investigate this problem, we examine reaching movements of the human arm in a crank rotation task. The experimental data show that after learning and accommodation the position, velocity, and the tangential force trajectories converge to unique profiles while that of the normal force do not. To explain the formation of the human trajectories observed in the experiments, we resort to the optimization approach. Different criteria of optimality are analyzed using simplified kinematic and dynamic models. From the kinematic analysis we suggested that in the process of learning and adaptation the brain constructs a suitable parameterization of the constraint manifold. In the dynamic analysis it is found that the human trajectories cannot not be captured by a single conventional criterion. It is shown that a weighted combination of the hand force changes and the joint torque changes produces better results but brings the problem of weight selection.

Many control strategies have been proposed to compensate for uncertainties of robotic system, where the uncertain parameters are assumed to be constant. However, this assumption may not be satisfied in practice. In this paper, we consider the tracking problem for robot manipulator with unknown and time-varying physical parameters and disturbances. Our proposed state feedback adaptive robust controller consists of a nonlinear compensation based on nominal physical parameters, a linear and a nonlinear state feedback, and an adaptation algorithm for adjustment of the gravitational parameters. The effects of time-varying parameters and disturbances on the tracking performance can be attenuated within a prescribed level. Exact asymptotic tracking can be achieved for set point control with vanishing disturbances and constant parameters. By adding feedback input to penalty variable, high gain feedback can be strategically avoided. The effectiveness of the proposed method is verified by simulation and experiments.

Adaptive resolution of function approximator is known to be important when we apply reinforcement learning to unknown problems. We propose to apply successive division and integration scheme of function approximation to temporal difference learning based on local curvature. TD learning in continuous state-space is based on non-constant values function approximation, which requires the simplicity of function approximator representation. We define bases and local complexity of function approximator in the similar way to the autonomous decentralized function approximation, but they are much simpler. The simplicity of approximator element bring us much less computation and easier analysis. The proposed function approximator is proven to be effective through function approximation problem and a reinforcement learning standard problem, pendulum swing-up task.

With the development of information science and robotic technology, it becomes more important to generate human interactive robots. The design platform for developing such robots should satisfy three basic conditions: (1) it can test safely the performance of the robot through the physical interaction with human, (2) human subject can estimate subjectively the outside appearance of the robot, and (3) it can simulate the dynamic human interactive robot motion within real-time. This paper proposes our immersion type dynamic simulation platform. An application to estimate the robots performance when performing cooperative object lifting task with human subject is chosen in order to show the effectiveness of our system. The analysis of the recorded data is useful to design the novel human interactive robots.

Kinematics aspect of planning and control of human arm movements are considered in this paper. First, we analyze unconstrained reaching movements using simple analytical models. Here, we introduce a generalized minimum jerk criterion, analyze the smoothness of the optimal solutions and discuss the role of the boundary conditions in the trajectory formation. Next, we analyze the constrained human movements using a crank rotation task. Analysis of the experimental data shows that constrained and unconstrained reaching movements might involve different control strategies, which reflected in the choice of the optimality criterion. It is demonstrated that among the kinematics based criteria, the minimum crank jerk criterion produce a rough matching to the experimental data. It is hypothesized that in the process of learning and adaptation the brain constructs a suitable parameterization of the strained manifold.

Human hand can not only catch and grasp the complex objects but also easily manipulate the objects by switching various types of interactions. Research on the basic mechanics and control principles of hand's dexterous manipulation abilities is one of the most important subjects in bio-mimetics. In this paper, we formalize the multi-fingered hand manipulation problem as a general dynamic complementarity (DC) system. Based on the fact that the DC system can be transformed to a mixed logical dynamical (MLD) model, and that MLD system can control problem can be solved using powerful mixed integer programming (MIP) algorithm, we propose to realize dexterous hand manipulations by using a cyber-grasp system, and study its robotic realization on GIFU hand III, which is equipped with tactile sensors. We suggest that in order to obtain sub-optimal solutions, biologically inspired techniques might be very powerful.

We consider the tracking problem for robot manipulators with unknown and time-varying physical parameters and disturbances which belong to L/sub 2/. Proposed state feedback adaptive robust controller consists of a nonlinear compensation based on nominal physical parameters, a linear and a nonlinear state feedback, and an adaptation algorithm for adjustment of the gravitational parameters. The effects of time-varying parameters and disturbances on the tracking performance can be attenuated within a prescribed level. Exact asymptotic tracking can be achieved for vanishing disturbances, constant parameters, and set point problem. By adding feedback input to penalty variable, high gain feedback can be strategically avoided.

An analysis of the stiffness of redundant manipulators is undertaken in this paper. First, the matrix of the force-dependent stiffness is derived and its basic properties are analyzed. In particular, in the planar case the stability conditions for the force dependent stiffness (and gravity-dependent stiffness) are obtained in the analytical form. Next, dual properties of the stiffness and compliance are exploited to establish a decomposition of the joint stiffness and compliance in the form similar to the decomposition of the joint velocities and torques. Finally, a minimal, nonredundant parameterization of the joint stiffness and compliance is commented.

This paper deals with a class of parallel mechanisms, called Gough-Stewart platforms. For these mechanisms, pre-loaded by driving forces, the stiffness matrix is derived and its basic properties are established. It is shown when the stiffness matrix becomes asymmetric, how the parametric imbalance may influence the system stability, and how the center of stiffness depends on the force pre-loading. Next, some necessary and some sufficient conditions for the stability are established in an analytical form by transforming the stiffness matrix to the center of stiffness.

Deals with a crank rotation task. The task requires coordinated movements of arm links without developing excessive internal forces. This research is directed, mainly, to the understanding of comfortable human movements constrained by the external environment. To get a deeper insight into the crank rotation task, we develop a mathematical model and analyze a weighted minimum norm muscle force distribution scheme that can be used in the resolution of the force redundancy. Analysis of experimental data shows that in comfortable motions a human is likely to modulate the rotational stiffness of the crank. This gives an additional constraint that can be used in the resolution of the force redundancy by optimization techniques.

This paper proposes a computational model for the cerebellum and spinal code. The proposed motion control scheme makes it possible to stabilize the posture of hand to the desired equilibrium trajectories. In this model, a motor neuron regulate the rest length of muscle to compensate the changes of equilibrium joint positions caused by the external load and the cerebellum compensates the inertial forces to stabilize the equilibrium position of joints. The asymptotic stability of joint angles to the desired equilibrium joint positions is proven theoretically and the performance of the control model is verified by some simple computer simulations.