10 years. What happened with this? Since we
have no database of stroke research and protocols this is lost to the
sands of time.
Renaud Ronsse
•
Tommaso Lenzi
•
Nicola Vitiello
•
Bram Koopman
•
Edwin van Asseldonk
•
Stefano Marco Maria De Rossi
•
Jesse van den Kieboom
•
Herman van der Kooij
•
Maria Chiara Carrozza
•
Auke Jan Ijspeert
Received: 23 February 2011/Accepted: 30 July 2011
Ó
International Federation for Medical and Biological Engineering 2011
Abstract
In this article, we propose a new method for providing assistance during cyclical movements. This method is trajectory-free, in the sense that it provides user assistance irrespective of the performed movement, and requires no other sensing than the assisting robot’s own encoders. The approach is based on adaptive oscillators,i.e., mathematical tools that are capable of learning the high level features (frequency, envelope, etc.) of a periodic input signal. Here we present two experiments that we recently conducted to validate our approach: a simple sinusoidal movement of the elbow, that we designed as a proof-of-concept, and a walking experiment. In both cases,we collected evidence illustrating that our approach indeed assisted healthy subjects during movement execution.Owing to the intrinsic periodicity of daily life movements involving the lower-limbs, we postulate that our approach holds promise for the design of innovative rehabilitation and assistance protocols for the lower-limb, requiring little to no user specific calibration.
1 Introduction
In modern robotics research, a lot of attention is devoted to service applications, with the general objective to improve the human daily life [39]. In particular, assistive and rehabilitation robots have been proposed as an innovative mean to improve the condition of people affected by chronic or momentary movement disabilities [9,13]. Assistive and rehabilitation robots have different goals.The former aims at assisting people affected by chronic movement disorders or neural lesions by providing continuous support giving extra power [20] or increasing movement accuracy [32]. On the other hand, the latter aims to retrain the nervous system and/or the musculoskeletal apparatus of the patient to restore his/her normal movement ability [27,45]. Despite of having different objectives, human robot interfacing is a critical issue for both assistive and rehabilitation robotics. The human-robot interface is indeed responsible for both power transfer and information transmission. More in detail, the human-robot physical interface is intended to provide a safe and comfortable interaction while transferring power between the two agents. Ergonomics studies [41] provide the main design guidelines for these interfaces, which are particularly critical in the case of wearable robots [7], due to the close interaction with the user. The human-robot cognitive interface instead is deputed to the acquisition and transfer of information regarding
the cognitive involvement of the patient in the task (e.g.,planning, reasoning, execution of a movement). In the case of assistive robots, the goal is typically to amplify the movement initiated by the user, sothat the effort spent by him/her is reduced without losing the control of the movement. On the contrary, rehabilitation robots exploit the information about the user intention to define the rehabilitative task in terms of spatiotemporal movement features.As such, the active participation of the patient in the task is promoted, and his/her effort is increased. This rehabilitative control strategy, commonly referred as ‘‘assist-as-needed’’,has been proved to be an effective way to increase the outcome of robot-mediated rehabilitation therapy by promoting motor recovery [1,11,12,45,54]. In this article, we describe a new approach that we recently developed to estimate the user’s intended movement while performing a cyclical motion task. This method can be used for both assistive and rehabilitative purposes.Unlike other methods previously used to estimate intended movements, our approach does not rely on inspecting activations by means of direct interfaces at the level of the central or peripheral nervous system or by electromyography (EMG) [7,21,22,36]. EMG based control has been successfully used to reduce the metabolic cost of walking of a healthy person [38], or to provide full-body daily assistance [20]. However, EMG recordings suffer from some drawbacks related to signal stability [6], which leads to the need of periodic recalibration and may also cause discomfort to the user over long periods of time (e.g., due to skin irritation). Our method requires no other sensing than the encoder of the robot actuators, avoiding the problems related to sensor placement, user-dependent calibration, or signal durability and reliability. As a consequence, our method provides both a fast and convenient integration to the user’s body and an adaptivity to the user’s intentions which—pending a sound and attractive ergonomic design—are the major requirements to maximize the device acceptability for potential users.To compensate for the ‘‘loss’’ of information that could have been provided by direct sensing of the user status (e.g.,EMGs), we embedded some apriori knowledge about the movement directly into the controller. In the case of lower-limb movements, this apriori knowledge simply consisted of assuming the movement to be periodic, a hallmark of daily life activities involving the lower-limbs (walking,running, stair climbing, etc.). The strategy proposed here exploits the concept of motor primitives, which emerged from biology [2,16] and has now clearly percolated in robotics [8,17]. The concept of motor primitives is very general in neuroscience, since motor primitives were identified at the cerebral, spinal, muscular, and kinematic levels.Nonetheless, the underlying idea of motor primitives is that a complex motor behavior can be described as the composition of simpler building blocks (i.e., the motor primitives) by using a finite set of parameters. The proposed movement estimation method follows this principle: Instead of directly estimating the intended movement kinematics(the epiphenomenon of the intended movement), we make use of the apriori knowledge that the movement is periodic to derive a non-linear dynamical system able to represent the movement in a finite set of simple features. Specifically, we make use of adaptive oscillators [4,29], a mathematical tool capable of synchronizing to a periodic signal and extracting its relevant features (like its frequency and envelope)through dynamical equations.In this article, we present the results of two recent experiments. Experiment 1 was conceived as a proof-of-concept of the whole approach. For that reason, we designed this experiment to be as simple as possible: we focused on sinusoidal movements about the elbow joint. As such, we avoided the intrinsic complexities related to the lower limb, like complex periodic joint profiles, multi joints coordination, and contacts with the ground. Nonetheless, we asked the participants to perform the movement around the vertical position, mimicking the inverted pendulum configuration of the leg during the stance phase of walking [15]. This experiment was already published in[31,33,34] and is only surveyed here. Experiment 2 extends the approach to walking assistance, and therefore specifically addresses the related challenges. Preliminary results were recently published in [35]. Both experiments deal with movement assistance of healthy participants.Therefore, we recorded biological signals—namely EMGs and oxygen consumption—illustrating that less effort (or energy) was required from the participants to perform the same movement, in steady-state regime. We further paid particular attention to design conditions illustrating the adaptive features of our controller, i.e., requiring the participants to modulate their limb trajectory. This last point is explored in conditions involving transient behavior, i.e.,changes in the movement pattern. Extension of our approach to rehabilitation protocols involving patients will be an intensive field for future research.
