A RLWPR network for learning the internal model of an anthropomorphic robot arm (original) (raw)
Related papers
Learning combined feedback and feedforward control of a musculoskeletal system
Biological Cybernetics, 1996
The goal of this paper is the learning of neuromuscular control, given the following necessary conditions: (1) time delays in the control loop, (2) non-linear muscle characteristics, (3) learning of feedforward and feedback control, (4) possibility of feedback gain modulation during a task. A control system and learning methodology that satisfy those conditions is given. The control system contains a neural network, comprising both feedforward and feedback control. The learning method is backpropagation through time with an explicit sensitivity model. Results will be given for a one degree of freedom arm with two muscles. Good control results are achieved which compare well with experimental data. Analysis of the controller shows that significant differences in controller characteristics are found if the loop delays are neglected. During a control task the system shows feedback gain modulation, similar to experimentally found reflex gain modulation during rapid voluntary contraction. If only limited feedback information is available to the controller the system learns to co-contract the antagonistic muscle pair. In this way joint stiffness increases and stable control is more easily maintained.
A proposed neural internal model control for robot manipulators
This paper describes the design of a Neural Internal Model Control (NIMC) system for robots, based on Recurrent Hybrid Networks (RHNs). The NIMC, an alternative to the basic inverse control scheme, consists of a forward internal neural model of robot, a neural controller and a conventional feedback controller. An Alopex Learning Algorithm (ALA) was used to adjust weights of the proposed neural network. Backpropagation (BP) algorithm is also employed for comparison. Diagonal Recurrent Networks (DRNs) and Feedforward Neural Networks (FNNs) controllers were used for comparison. The robot in this study was adept one SCARA type robot manipulator.
A Proposed Neural Internal Model Control for Robot Manipulator
2013
This paper describes the design of a Neural Internal Model Control (NIMC) system for robots, based on Recurrent Hybrid Networks (RHNs). The NIMC, an alternative to the basic inverse control scheme, consists of a forward internal neural model of robot, a neural controller and a conventional feedback controller. An Alopex Learning Algorithm (ALA) was used to adjust weights of the proposed neural network. Backpropagation (BP) algorithm is also employed for comparison. Diagonal Recurrent Networks (DRNs) and Feedforward Neural Networks (FNNs) controllers were used for comparison. The robot in this study was adept one SCARA type robot manipulator.
Design of a neural internal model control system for a robot
Robotica, 2000
This paper describes the design of a Neural Internal Model Control (NIMC) system for robots, based on Recurrent Hybrid Networks (RHNs). The NIMC, an alternative to the basic inverse control scheme, consists of a forward internal neural model of robot, a neural controller and a conventional feedback controller. An Alopex Learning Algorithm (ALA) was used to adjust weights of the proposed neural network. Backpropagation (BP) algorithm is also employed for comparison. Diagonal Recurrent Networks (DRNs) and Feedforward Neural Networks (FNNs) controllers were used for comparison. The robot in this study was adept one SCARA type robot manipulator.
N90-29030 "Computational" Neural Learning Formalisms for Manipulator Inverse Kinematics
An efficient, adaptive neural learning paradigm for addressing the inverse kinematics of redundant manipu-lators is presented. The proposed methodology exploits the infinite local stability of terminal attractors -a new class of mathematical constructs which provide unique information processing capabilities to artificial neural systems. For robotic applications, synaptic elements of such networks can rapidly acquire the kine-matic invariances embedded within the presented samples. Subsequently, joint-space configurations, required to follow arbitrary end-effector trajectories, can readily be computed. In a significant departure from prior neuromorphic learning algorithms, this methodology provides mechanisms for incorporating an in-training "skew" to handle kinematics and environmental constraints.
A Neural Network System with Reinforcement Learning to Control a Dynamic Arm Model
Biomedical Engineering: Applications, Basis and Communications, 2001
A neural network system is presented for controlling a two-link dynamic arm model where the task is to move the arm from any initial position to any final position in the sagittal plane. The controller produces joint torque-time profiles that begin and end with equilibrium values at the initial and final positions, respectively. A memory type neural network is trained by supervised learning methods to predict the joint's static equilibrium torque values corresponding to joint angles. A reinforcement learning network is used to determine the parameters needed for synthesis of the torquetime profiles for each joint. The reinforcement signal is computed based on the distance between the desired end point position and velocity and the states achieved based on the generated torque profiles. The general pattern of the torque-time plots is decided a priori according to the literature. The methods of training and an illustrative example of the algorithm's performance are presented.
Human Motor Computational Model Through Iterative Model Reference Adaptive Control
2011
A computational model using mechanical impedance control in combination with an iterative model reference adaptive control is proposed to capture the hypothesised mechanisms in human motor control and the learning capacity that humans exhibit in adapting their movements to new and unstructured environments. The model uses an iterative learning control law to model human learning through repetitive processes. In this proposed framework, motion command is carried out without the need for inverse kinematics. Learning is performed without an explicit internal model of the body or of the environment. The resulting framework is simulated and compared to the experimental data, involving subjects performing a task in the face of specified disturbance forces. In this paper, a stable task is specifically addressed as the example. It is shown that the proposed framework produces a stable and accurate description of the general behaviour observed in the human motor adaptation.
Human motor learning through iterativemodel reference adaptive control
2011
A computational model using mechanical impedance control in combination with an iterative model reference adaptive control is proposed to capture the hypothesised mechanisms in human motor control and the learning capacity that humans exhibit in adapting their movements to new and unstructured environments. The model uses an iterative learning control law to model human learning through repetitive processes. In this proposed framework, motion command is carried out without the need for inverse kinematics. Learning is performed without an explicit internal model of the body or of the environment. The resulting framework is simulated and compared to the experimental data, involving subjects performing a task in the face of specified disturbance forces. In this paper, a stable task is specifically addressed as the example. It is shown that the proposed framework produces a stable and accurate description of the general behaviour observed in the human motor adaptation.
Modular Neural Net System for Inverse Kinematics Learning
International Conference on Robotics and Automation, 2000
Inverse kinematics computation using an artificialneural network that learns the inverse kinematics ofa robot arm has been employed by many researchers.However, conventional learning methodologies do notpay enough attention to the discontinuity of the inversekinematics system of typical robot arms with joint limits.The inverse kinematics system of the robot armsis a multi-valued and discontinuous function. Since itis difficult for a well-known
2005
Inverse kinematics computation using an artificial neural network that learns the inverse kinematics of a robot arm has been employed by many researchers. However, the inverse kinematics system of typical robot arms with joint limits is a multi-valued and discontinuous function. Since it is difficult for a well-known multi-layer neural network to approximate this kind of function, a correct inverse kinematics model cannot be obtained by using a single neural network. To overcome the difficulties of inverse kinematics learning, we proposed a novel modular neural network system that consists of a number of experts, with each expert approximating a continuous part of the inverse kinematics function. The proposed system selects one appropriate expert whose output minimizes the expected position/orientation error of the end-effector of the arm. The system can learn a precise inverse kinematics model of a robotic arm with equal or more degrees of freedom than that of its endeffector. However, there are robotic arms with fewer degrees of freedom. The system cannot learn a precise inverse kinematics model of this kind of arm. To overcome this, we adopted a modified Gauss-Newton method for finding the least-squares solution. Through the modifications presented in this paper, the improved modular neural network system can obtain a precise inverse kinematics model of a general robotic arm.