Shou-Han Zhou

Learning control in robot-assisted rehabilitation of motor skills – a review

The key idea in iterative learning control is captured by the intuition of ‘practice makes perfect’. The underlying learning is based on a gradient descent algorithm iteratively optimising an appropriate input–output measured criterion. How this paradigm is used to model quantitatively, at an input/output level, the learning that happens in the context of human motor skill learning is discussed in this note. Experimental studies of human motor learning, in robotically controlled environments, indicate that a model consisting of a classical (iterative) learning control augmented with an appropriate kinematic model of human motor motion fits the observed human learning behaviour well. In the context of the rehabilitation of motor skills, such models promise better human–machine interfaces that extend the capability and capacity of rehabilitation clinicians by creating effective robot–patient–clinician feedback loops. The economic promise of robot-assisted rehabilitation is to greatly extend the intervention...

Modeling Individual Human Motor Behavior Through Model Reference Iterative Learning Control

A computational model is proposed in this paper to capture learning capacity of a human subject adapting his or her movements in novel dynamics. The model uses an iterative learning control algorithm to represent human learning through repetitive processes. The control law performs adaptation using a model designed using experimental data captured from the natural behavior of the individual of interest. The control signals are used by a model of the body to produced motion without the need of inverse kinematics. The resulting motion behavior is validated against experimental data. This new technique yields the capability of subject-specific modeling of the motor function, with the potential to explain individual behavior in physical rehabilitation.

Point-to-point learning in human motor systems

Motor planning is one of the key processes in human motor systems. It accounts for how the Central Nervous System (CNS) generates the desired trajectory for the human to follow in order to achieve the given task. This work focuses on a special task: point-to-point tracking. For this task, point-to-point learning is used with the aim for gaining a deeper understanding of motor learning. Point-to-point learning control (LC) not only naturally matches the task, but also provides tools to analyse learning in human motion planning. It is observed that the simulation results obtained by using tracking performance failed to match the observed motion performed by humans captured from experiments. It was then shown that additional objective, namely minimal jerk, is required to provide satisfactory results, suggesting that humans consider objectives other than tracking during motion planning. This finding provides experimental and mathematical validation to current hypotheses of human motor planning.

Effect of sensory experience on motor learning strategy

It is well known that the central nervous system automatically reduces a mismatch in the visuomotor coordination. Can the underlying learning strategy be modified by environmental factors or a subjects learning experiences? To elucidate this matter, two groups of subjects learned to execute reaching arm movements in environments with task-irrelevant visual cues. However, one group had previous experience of learning these movements using task-relevant visual cues. The results demonstrate that the two groups used different learning strategies for the same visual environment and that the learning strategy was influenced by prior learning experience.

Human Motor Learning Through Iterative Model Reference Adaptive Control

Abstract 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.

On on-line sampled-data optimal learning for dynamic systems with uncertainties

In this study, a novel on-line optimal learning control is proposed to achieve the optimal performance for dynamic systems with modeling uncertainties, measurement noise and iteration-varying initial conditions. By introducing a nominal model and a sampled-data controller, it is possible to find the optimal solution iteratively of an optimization problem using gradient descent method. A feedback controller is introduced along the finite-time domain to ensure that the difference between the output of the nominal model and that of the actual plant can be made arbitrarily small. This feedback thus can be used to handle various uncertainties in the plant model, while the feedforward learning controller is used to ensure the convergence of the plant output to the optimal solution. Hence, by tuning sampling period and feedback gain matrix, it is possible to ensure that the output of plant converges semi-globally practically to the optimal solution. Simulation results illustrate the effectiveness of the proposed method.

Modelling of human motor control in an unstable task through operational space formulation

Human motor control computational model is an important component in the study and the successful realisation of human-robot interaction. In this paper, the Operational Space Formulation is presented as a suitable framework of human motor control computational model based on the Equilibrium Point Hypothesis (EPH) approach. The iterative adaptive control strategy was incorporated to simulate human motor adaptation to different tasks. The strategy involves the use of an Equilibrium Model which represents the ideal human motor response to a given task. The combined strategy was simulated to match a set of data gathered experimentally from several human subjects. The results were observed to explain many of the features found in the recorded behaviours in the EPH-based approach of human motor modelling.

Modeling of Endpoint Feedback Learning Implemented Through Point-to-Point Learning Control

In the last decade, several experiments were conducted to investigate human motor control behavior for the task of arm reaching, using only visual feedback of the final hand position at the end of each reaching motion. Current computational frameworks have yet to model that the humans learn to complete such a task by feedforward action based on the feedback of a displacement error at the end of past reaching motions. This paper demonstrates how such learning can be formulated as an optimization problem. By designing a cost function which weighs the tracking of the target and the smoothness of human motion, the constructed framework, implemented in the form of point-to-point learning control, inherently embeds the feedforward control and enables learning over repeated trials using only the available feedback from past observations, here the endpoint errors of a reaching motion trajectory. The proposed framework is able to reproduce the human learning behavior observed in experiments.

On experimentally validated iterative learning control in human motor systems

A framework is developed to construct computational models of the human motor system (HMS) using iterative learning control (ILC) update structures. Optimal models of movement are introduced using a cost function that is motivated by learned human motion results. Three general ILC update structures are derived that each generate the required limiting solution using different forms of experimental data. It is shown how the parameters in each that govern convergence permit varying degrees of freedom in capturing the observed learning transients. Experimental results in which a participant uses a planar robot to perform reaching tasks confirm the ability of the proposed ILC structures to accurately model the learning ability of the human motor system.

Trajectory redundancy iterative learning control

For tasks which require a robot to track some particular points along a trajectory (instead of the whole trajectory), there exists redundancy. This redundancy results in an increase in the feasibility in the controller design, enabling the possibility of the robot to obtain better performance by satisfying secondary objectives whilst performing the primary objective of tracking the target points. This paper addresses the task redundancy by using point-to-point learning control. It is shown to be an effective tool to accommodate trajectory redundancy since it has the ability to fully explore the increased feasibility resulting from such redundancy. Following the similar idea widely used in kinematic redundancy, a decomposition technique is used. This leads to a simplification of constrained optimization and provides a suboptimal performance in terms of secondary task while the primary task is always achieved. As an example, the formulation is implemented in an on-line fashion to enable a non-redundant robot to track a target point whilst avoiding an obstacle. Simulation results shows good performance from the proposed online algorithms.