In this paper we survey a number of methods for teaching human motion/force skills to robots. Since verbal description of a human skill could only provide qualitative guidelines for machine programming, several works were motivated to identify human skills analytically through the study of human teaching data. The key advantage of direct teaching by human demonstration is that the method is highly user-friendly; it requires no special knowledge of machine programming from the users, and it automatically extracts user intentions and strategies from task performance demonstrated by the users. From the early teaching/playback method to the recent methods of teaching sensory feedback laws, a historical perspective of the field of human skill teaching is given in this paper. Limitations and applicability of methods under review are discussed. Essential issues such as representation and identification of human skills as well as consistency of teaching information are addressed.
This paper proposes a new control algorithm of servo-based impedance control of robot manipulator for stabilization of contact task. There still has been stability problem of servo-based impedance control when a manipulator contacts to a stiff environment with a small viscosity coefficient of target impedance. The paper shows a stabilizing method for impedance control even with a small target viscosity. The proposed control system modifies the target dynamics model, but has no necessity to change the servo-control system. The simulations and contact task experiments show an extremely similar response to a target impedance and verify sophisticated contact stability with an one-tenth target viscosity and with over two times larger velocity more than the stability conditions limited for conventional control methods.
Robots are required to have compliant joints to decrease a collision force in coming into contact with an object to be grasped or the environment. In this point, an artificial rubber muscle called Rubbertuator is promising for applications to contacting tasks, because its spring characteristics are useful to absorb the collision force. In this paper, we suggest how to control joint stiffness and torque independently by applying an impedance control to generate a required air pressure in respective two Rubbertuators which actuate a joint antagonistically. Besides, the sliding mode control, which can guarantee robust control performances against the Rubbertuator's inherent uncertain characteristics, is applied to the force control system of the Rubbertuator. Experimental results show the feasibility of proposed control system.
Nonholonomic constraints are exploited to design a controllable n-joint manipulator with only two inputs. Gears subject to nonholonomic constraints are designed to transmit velocities from the inputs to the unactuated joints. The designed nonholonomic manipulator is shown to be completely controllable in the whole configuration space. The system possesses a triangular structure for which a conversion into chained form is presented. The nonholonomic manipulator can, therefore, be controlled with open loop or closed loop using existing controllers for chained form. Simulation results are presented showing steering between two configurations in the n-dimensional joint space.
We present an efficient method to fixate a binocular gaze point on an object moving around in a complicated environment. Assuming the object is in the vicinity of the gaze point, the image features of the object can be isolated from others by using zero disparity filtering (ZDF). The object position in the depth direction, however, can not be estimated from such isolated features. In order to get the depth information simultaneously with the isolation by ZDF processing, we introduce the novel localization technique based on the idea of “virtual horopter”. The proposed method is implemented on our active vision system with much reduced programming efforts and resources in comparing with conventional methods. Several experiments demonstrate that the binocular tracking based on the proposed method works for various shape objects and in complicated environments.
Recently the use of neural networks as the inverse-kinematics model of a robot arm has been proposed in learning control of the robot arm. The forward and inverse modeling, the feedback error learning schema and the goal directed model inversion were proposed to extend the acquisition of the inverse model for the systems with many-to-one inputoutput correspondence. However, these methods can be used only for on-line learning. The learning of neural networks usually requires many iterations of robot arm movement and of its position measurement. In order to reduce the number of movements of the robot arm, the hybrid system which consists of a learning element and an extended feedback controller are proposed. The learning element approximates the inverse kinematics model of the robot arm. By using the extended feedback controller, the high precision solutions of the inverse-kinematics problems are obtained so that these solutions can be used for the teaching signal of the learning element. After the acquisition of these solutions, the off-line learning of the learning element is conducted. The use of forward model of the robot arm is also proposed. The numerical simulations show the good performance of the proposed system.
A novel ultrasonic sensing method which can measure the normal direction of walls is proposed. The inclination of the wall is calculated from the difference of the round-trip time-of-flight detected by using plural receivers. This sensor system has four receivers for the error correction using inconsistency check of the propagation times.
This paper presents an efficient control approach for redundant manipulators using a torque-based formulation. First, the manipulator with redundant degrees of freedom is decomposed into several non-redundant subsystems. Then, each of these subsystems is analyzed for the conditioning of the respective Jacobian matrix, considering the dynamic performance. Thus a set of possible solutions is constructed. Further on, the formulation of the proposed redundancy resolution control is realized using the relation between the joint torque and the given end-effector acceleration (torque-based formulation). The approach is suitable for parallel processing, and real-time dynamic redundancy control of redundant manipulators is possible. Numerical simulations have been performed to show the effectiveness.
In this paper, a compensation method of disturbances using a disturbance observer is proposed for a force control of a pneumatic robot manipulator. The generated torque by a pneumatic actuator can be estimated based on the pressure signals. The inner torque control system is constructed by feeding back the generated torque to improve the dynamic characteristics of the actuator. In order to reduce the influence of disturbances comprising friction torque, parameter variations of plant and environment and so on, the reaction torque control system is constructed with a disturbance observer which estimates the disturbances based on the reference input to the inner torque control system and the reaction torque sensed with a force sensor. Further, to control the reaction torque without a force sensor, we construct the reaction torque estimation observer using the generated torque, the nominal friction torque and the angular velocity. In the above reaction torque control system, the estimated reaction torque is used instead of the sensed torque, so that the force sensorless torque control system can be constructed. From some simulations and experiments, it is confirmed that the proposed control system is effective to improve the robustness for the friction torque and the parameter change of object in the force control of a pneumatic robot manipulator.
The mobile robots which can walk around and perform several tasks over 3D terrain, a generalized terrain including the surface of wall and ceiling of large constructions are highly demanded. We discuss the control of the leg motion of quadruped-wall-climbing-robot specifically designed for 3D terrain from the view point of two design and control concepts which we have already proposed, the GDA, or gravitationally decoupled actuation, and Coupled Drive. The GDA was introduced to eliminate negative power consumption and improve energy efficiency in ground walking, and the Coupled Drive was introduced to evenly distribute power generation among installed actuators, decrease the weight of the actuation mechanism, and enable powerful walking in wall climbing motion. In this article we propose a new control method to adaptively select walking postures in both ground walk and wall climbing and optimize the walking performances in terms of the concepts of GDA and Coupled Drive. We made simulation experiments considering actual actuation characteristics, standing motion constraint and other boundary conditions and shows that the introduced control improve walking performance in 3D terrain well.
When the size of distributed robotic system becomes large, local communication system is considered appropriate for cooperation in the light of cost and capacity. The behavior of robots has a respectable effect on the efficiency of local communication system. When the number of robots required for cooperative tasks is given, it is therefore essential to know what kind of behavior robots should take so that information is passed to those robots efficiently. Group behavior is considered to have advantages in improvement of communication performance, as previous researches show its effectiveness in such tasks as sample collecting or searching. This paper analyzes the effect of group behavior on the performance of local communication system of multiple mobile robots. We derive a set of differential equations that describe the relationship between group size and the information diffusion among robots. The analytical results show the effectiveness of group behavior and allow us to calculate the optimal group size that minimizes the diffusion time for desired number of robots, even if the number is stochastically distributed. The validity of the analysis is verified by computer simulation.