Anticipatory assistance-as-needed control algorithm for a multijoint upper limb robotic orthosis in physical neurorehabilitation

https://doi.org/10.1016/j.eswa.2013.11.047Get rights and content

Highlights

  • We propose a new anticipatory assistance-as-needed robotic control algorithm.

  • The control algorithm adapts its behavior to the patient dysfunctional profile.

  • The algorithm anticipates to trajectory deviations to encourage neural plasticity.

  • We validated the proposed algorithm throughout a robotic simulator.

Abstract

Robotic devices are becoming a popular alternative to the traditional physical therapy as a mean to enhance functional recovery after stroke; they offer more intensive practice opportunities without increasing time spent on supervision by the treating therapist. An ideal behavior for these systems would consist in emulating real therapists by providing anticipated force feedback to the patients in order to encourage and modulate neural plasticity. However, nowadays there are no systems able to work in an anticipatory fashion. For this reason, the authors propose an anticipatory assistance-as-needed control algorithm for a multijoint robotic orthosis to be used in physical ABI neurorehabilitation. This control algorithm, based on a dysfunctional-adapted biomechanical prediction subsystem, is able to avoid patient trajectory deviations by providing them with anticipatory force-feedback. The system has been validated by means of a robotic simulator.

Obtained results demonstrate through simulations that the proposed assistance-as-needed control algorithm is able to provide anticipatory actuation to the patients, avoiding trajectory deviations and tending to minimize the degree of actuation. Thus, the main novelty and contribution of this work is the anticipatory nature of the proposed assistance-as-needed control algorithm, that breaks with the current robotic control strategies by not waiting for the trajectory deviations to take place. This new actuation paradigm avoids patient slacking and increases both participation and muscle activity in such a way that neural plasticity is encouraged and modulated to reinforce motor recovery.

Introduction

ABI (Acquired Brain Injury) is defined as an injury to the brain that has occurred after birth but it is not related to congenital defects or degenerative diseases (Brain injury association of america, 2012). The WHO (World Health Organization) estimated that in 2005, stroke accounted for 5.7 million deaths worldwide, equivalent to 9.9% of all deaths, and it was the main cause of disability, afflicting 30.7 million people (World health organization, 2012). These days, nine million people suffer from a cerebrovascular disease every year in the world (World health organization, 2012) and globally, stroke is the second leading cause of death and the eighth cause of severe disability in the elderly. By the year 2020, as the WHO predicts, it will be among the ten most common causes of disability in the developed world. These injuries, due to their physical, sensory, cognitive, emotional and socio-economic consequences, considerably change the life of both the patients and their families. The cause of ABI can be either traumatic (car accidents, falls, etc.) or non-traumatic (strokes, brain tumors, infections, etc.). The most common ABIs are stroke and TBI (Traumatic Brain Injury) (Murray & Lopez, 1997).

New techniques of early intervention and the development of intensive ABI care have noticeably improved the survival rate (The internet stroke center, 2012). However, in spite of these advances, brain injuries still have no surgical or pharmacological treatment to re-establish lost function. Neurorehabilitation therapies address this problem by restoring, minimizing or compensating the functional alterations in people with disabilities of neurological origin. Medical evidence in neurorehabilitation is scarce and the assessment methods, especially those dealing with upper limb function, depend on clinician experience and subjectivity. Moreover, motion analysis assessments, which are more sensitive and provide objective data, are mainly centered on gait analysis, whereas upper limb tests are still not widely performed; current trend in development towards individualised and more complex models needs to be justified by demonstrating their ability to answer questions that cannot already be answered by existing models (Bolsterlee, Veejer, & Chadwick, 2013). Besides, the lack of standardized protocols due to the large variety of movements, complexity of the upper extremity and lack of international consensus to validate the protocols hampered the advance on this area (van Andel, Wolterbeek, Doorenbosch, Veeger, & Harlaar, 2008).

One of the main objectives of neurorehabilitation is to provide patients with the capacity to perform specific ADL (Activity of the Daily Life) required for an independent life, taking into account that continual practice of fundamentally inappropriate compensatory strategies may be a critical factor limiting recovery after brain damage (Carr and Shepherd, 1989, Davies, 1990). Although traditional physical therapy can enhance functional recovery after stroke, robotic devices may offer more intensive practice opportunities without increasing time spent on supervision by the treating therapist (Dobkin, 2004). This, along with the assertion that traditional therapies are expensive and likely dosage dependant, have caused a remarkable increase in research aimed at creating, controlling and using robotic devices (Conesa et al., 2012, Wolbrecht et al., 2008).

Robotic neurorehabilitation is attractive because of its potential for easy deployment, its applicability across a wide range of motor impairment and its high measurement reliability and thus, there is an increasing interest in using these devices to support neurorehabilitation therapies (Riener, Nef, & Colombo, 2005). Moreover, it is also believed that robotic therapy during the acute and sub-acute phase of stroke recovery could augment changes in impairment driven by spontaneous biological recovery processes (Huang & Krakauer, 2009).

To provide patients with ADL-based functional rehabilitation under the assistance-as-needed paradigm (Emken, Bobrow, & Reinkensmeyer, 2005) (which means to assist the subject only as much as is needed to accomplish the task) and without the presence of a therapist but under his/her supervision, is one of the main challenges of the current neurorehabilitation technologies. Current assistance-as-needed strategies face one crucial challenge: the adequate definition of the desired limb trajectories regarding space and time that the robot must generate to assist the user during the exercise (Belda-Lois et al., 2011).

Rehabilitation robotic control algorithms can be grouped according to the strategy taken to facilitate motor recovery: assisting, challenge based, haptic simulation and non-contact coaching (Marchal-Crespo & Reinkensmeyer, 2009). Assistive controllers actively help the patients to achieve certain goals; challenge-based ones provide resistance to the performed movements. Haptic simulation consists in practising ADL movements in virtual environments. Coaching robotic systems do not physically interact with the patients but provide them with help and motivation.

Besides, there is a scientific theory, called the “Slacking Hypothesis”, that suggests that active guidance may decrease motor learning because, in some cases, it can cause patients to decrease their own effort during the training session (Wolbrecht et al., 2007). Thus, assistance-as-needed neurorehabilitation paradigm, which consists in providing the patients only with the assistance they need to perform certain activity, appears as a strong alternative to enhance the therapy outcomes. This actuation paradigm has been proven to be successful in previous motor rehabilitation studies (Barbeau & Visintin, 2003).

Several approaches to the assistance-as-needed paradigm can be found in the scientific literature. Some robotic systems provide an assistance that is proportional to the deviation of the patient given a predefined trajectory. Well known examples of this control strategy are MIT-MANUS (Krebs et al., 1998, Krebs et al., 2003, Krebs and Volpe, 2013), MIME (Lum et al., 1995, Lum et al., 2002, Lum et al., 2006), GENTLE/G (Loureiro & Harwin, 2007), ARMin (Nef et al., 2007, Gijbels et al., 2011, Guidali et al., 2011), L-EXOS (Montagner et al., 2007, Frisoli et al., 2012), ReoGo (Bovolenta, Sale, Dall’Armi, Clerici, & Franceschini, 2011) or NeReBot (Rosati, Gallina, & Masiero, 2007). Other systems that apply the aforementioned control strategy are also (Denve et al., 2008, Hesse et al., 2003, Richardson et al., 2006, Toth et al., 2005, Tsagarakis and Caldwell, 2003). These assistive robotic therapy controllers focus on the following idea: when the subject moves along a desired trajectory (and an artificially created virtual tunnel), the robot should not intervene, and if the participant deviates from the desired trajectory, the robot must create a restoring force (Marchal-Crespo & Reinkensmeyer, 2009).

Dynamic control systems, that are able to adapt to the current needs of the patient based on online performance measurements, can be also found in the scientific literature. The basis of these control strategies is to adapt their configuration parameters tuning the system to the subject changing needs. Riener et al. (2005) developed such system for gait rehabilitation by recognizing the patient intention and adapting the level of assistance to the subject’s contribution. Regarding the upper limb, inter-session parameter adaptation methods that allow the selection of the working parameters once a previous performance measurement is available can be found (Krebs et al., 2003, Kahn et al., 2004). Recently, Guidali et al. developed a method that made the robotic device able to react in real time to the performance of the subject by updating a dynamic model of the upper limb (Guidali, Schlink, Duschau-Wicke, & Riener, 2011); even though this work supposes a clear step forward to the work presented by Wolbrecht et al. (2008) (whose method was movement-specific) their ‘assistance-as-needed’ strategy is not focused on the provision of anticipatory force-feedback to the patients, in contrast, their aim is to perform an online adaptation of the amount of support depending on the activity. Finally, some assistance strategies introduce a forgetting factor to keep a challenging assistance level for the patient in order to avoid slacking (Emken et al., 2005, Guidali et al., 2011, Mihelj et al., 2007, Wolbrecht et al., 2007).

Anticipatory control is still a relatively unexplored niche in the field of rehabilitation robotics. No works have been found that try to anticipate patient intention in order to avoid trajectory deviations. However it is worth to mention the work developed by Everarts, Vallery, Bolliger, and Ronsse (2013), who proposed an anticipatoty algorithm to enhance robotic transparency for gait rehabilitation taking advantage of the cyclic nature of the gait; in this work a predictive layer is incorporated to the control architecture to compensate the computational delays, the mechanical response of the robot and the limited bandwidth.

In relation with intention detection, there are several robotic control mechanisms that rely on the information provided by EMG (Electromyography) signals (Lenzi et al., 2012, Song et al., 2013). In these works, based on single DoF (Degree of Freedom) orthoses, patient intention is detected in such a way that the motion is supported, saving physical effort but leaving full control to the subject; as a consequence, potential trajectory deviations are not corrected. Other authors, with the hypothesis that motor learning is stimulated by correlating motor commands with feedback signals to the somatosensory cortex, propose control schemas based on non-invasive BCIs (Brain Computer Interfaces) to restore lost motor function by routering movement related signals from the brain to external effectors (Birbaumer and Cohen, 2007, Birbaumer et al., 2009, Hayashi et al., 2012, Hochberg et al., 2006); the use of these interfaces are still at a very early stage of development and need further experimentation (Birbaumer et al., 2009).

Nowadays, there are not many robotic devices specifically oriented for practising ADLs, being the ADLER orthosis (Johnson et al., 2007) the most relevant amongst them. This device uses the HapticMaster (Van der Linde, Lammertse, Frederiksen, & Ruiter, 2002) robot to assist the patient along programmed ADL trajectories providing customized forces through three active DoFs (the other three remain passive). ADLER system is able to work under both assistive and challenge-based strategies. A limiting factor of this device is its small range of motion.

Full and more exhaustive reviews of rehabilitation robotics control strategies can be found in Marchal-Crespo and Reinkensmeyer (2009) and Loureiro, Harwin, Nagai, and Johnson (2011).

As it can be noticed by reading previously cited bibliography, although several adaptive methods have been developed, there are no upper limb-centered control algorithms intended to anticipate the patients in order to avoid deviations from those considered as healthy motions. The main goal of this research work is the design and definition of an anticipatory assistance-as-needed control algorithm to emulate, throughout a robotic orthosis, a therapist that is in direct contact with the patient when he/she is carrying out an ADL-based neurorehabilitation session. This algorithm is based in the fact that healthy individuals continuously select and weight different proprioceptive feedback to adjust their motion (Dietz, 2009), describing an anticipatory motor control for the production of movement (Pollok, Gross, Kamp, & Schnitzler, 2008). The anticipatory component of the algorithm refers to the ability to predict the trajectory of the patient to online adapt the response.

In this way, the major novelty and contribution of this research work is the breaking with the traditional control strategies: from reactive systems to an anticipatory system that takes into account the patient’s dysfunctional profile.

Section snippets

Methods

The used biomechanical model is an extended version of that one previously used by the authors in Pérez et al., 2010, Pérez-Rodríguez et al., 2012, with eight DoFs. Human upper limb motion is approximated as the articulated motion of rigid body parts (Biryukova, Roby-Brami, Frolov, & Mokhtari, 2000): scapula (from the clavicle to the shoulder joint), upper arm (between the shoulder and elbow joints), forearm (between the elbow and wrist joints) and hand (from the wrist joint on). For this field

Results and discussion

Table 7 shows the mean assistance percentage that the system has provided to the healthy subjects working with the standard configuration in the performed simulations. If these values are compared with those ones presented by Table 8, that contains the assistance percentage given to each of the pathological subjects that participated in this study, it can be clearly seen that the assistance-as-needed control algorithm almost do not assist healthy subjects while it behaves completely different

Conclusions

In this paper, the authors present an assistance-as-needed control algorithm able to provide ABI patients with anticipatory force-feedback only when it is strictly needed, avoiding the subjects to slack by increasing their participation and muscle activity in such a way that neural plasticity is encouraged and modulated to reinforce motor recovery. Although the algorithm presented here are designed and presented using the simulator, the realism that have been given to the developed robotic

Author’s contributions

RPR contributed in the research, design and validation of the proposed control algorithm and collaborated with UC in the interpretation of the obtained results. UC contributed with the data acquisition and worked together with RPR in the data formalization and modeling process. CR contributed in the development of the orthosis simulator. RPR, CC, JMT, JM and EJG contributed with the original study design.

Finally, all authors participated in the drafting of the manuscript and approved its final

Acknowledgements

This research work was partially funded by CDTI (project: REHABILITA; CIN/1559/2009), Spanish Government. The authors would like to thank all the REHABILITA consortium members, project ECNI-Estimulación Cerebral Invasiva y Rehabilitación asistida por robots para acelerar la rehabilitación en TCE, Instituto de Salud Carlos III, Ministry of Science and Innovation-PI082004, project 3e+D and ACC10 (Department of Industry, Generalitat de Catalunya).

References (83)

  • R. Miall et al.

    Forward models for physiological motor control

    Neural Networks

    (1996)
  • I.A. Murray et al.

    A study of the external forces and moments at sholuder and elbow while performing every day tasks

    Clinical Biomechanics

    (2004)
  • C. Murray et al.

    Alternative projections of mortality and disability by cause 1990–2020: Global burden of disease study

    The Lancet

    (1997)
  • R. Pérez-Rodríguez et al.

    Inverse kinematics of a 6 DoF human upper limb using ANFIS and ANN for anticipatory actuation in ADL-based physical neurorehabilitation

    Expert Systems with Applications

    (2012)
  • R. Shadmehr

    Computational approaches to motor control

  • C. van Andel et al.

    Complete 3D kinematics of upper extremity functional tasks

    Gait & Posture

    (2008)
  • D. Wolpert

    Computational approaches to motor control

    Trends in Cognitive Sciences

    (1997)
  • D. Wolpert

    Internal models in the cerebellum

    Trends in Cognitive Sciences

    (1998)
  • D. Wolpert et al.

    Multiple paired forward and inverse models for motor control

    Neural Networks

    (1998)
  • D. Ashby et al.

    Evidence-based medicine as bayesian decision-making

    Statistics in Medicine

    (2000)
  • J.M. Belda-Lois et al.

    Rehabilitation of gait after stroke: A review towards a top-down approach

    Journal of NeuroEngineering and Rehabilitation

    (2011)
  • N. Birbaumer et al.

    Brain-computer interfaces: Communication and restoration of movement in paralysis

    The Journal of Physiology

    (2007)
  • B. Bolsterlee et al.

    Clinical applications of musculoskeletal modelling for the shoulder and upper limb

    Medical and Biological Engineering and Computing

    (2013)
  • F. Bovolenta et al.

    Robot-aided therapy for upper limbs in patients with stroke-related lesions. Brief report of a clinical experience

    Journal of NeuroEngineering and Rehabilitation

    (2011)
  • Brain injury association of america. (2012)....
  • BTS Bioengineering. (2012)....
  • M. Cirstea et al.

    Compensatory strategies for reaching in stroke

    Brain

    (2000)
  • L. Conesa et al.

    An observational report of intensive robotic and manual gait training in sub-acute stroke

    Journal of NeuroEngineering and Rehabilitation

    (2012)
  • P. Corke

    Robotics, vision & control: Fundamental algorithms in Matlab

    (2011)
  • Costa, U., Opisso, E., Pérez, R., Tormos, J. M., & Medina, J. (2010). 3D motion analisys of activities of daily living:...
  • L.M. Crespo et al.

    Haptic guidance can enhance motor learning of a steering task

    Journal of Motor Behavior

    (2008)
  • P. Davies

    Right in the middle: Selective trunk activity in the treatment of adult hemiplegia

    (1990)
  • J. Denavit et al.

    A kinematic notation for lower-pair mechanisms based on matrices

    Transactions of ASME

    (1955)
  • A. Denve et al.

    Control system design of a 3-DOF upper limbs rehabilitation robot

    Computer Methods and Programs in Biomedicine

    (2008)
  • Emken, J., Bobrow, J., & Reinkensmeyer, D. (2005). Robotic movement training as an optimization problem: Designing a...
  • Everarts, C., Vallery, H., Bolliger, M., & Ronsse, R. (2013). Adaptive position anticipation in a support robot for...
  • T. Flash et al.

    The coordination of arm movements: An experimentally confirmed mathematical model

    The Journal of Neuroscience

    (1985)
  • A. Frisoli et al.

    Rehabilitation training and evaluation with the l-exos in chronic stroke

  • D. Gijbels et al.

    The armeo spring as training tool to improve upper limb functionality in multiple sclerosis: A pilot study

    Journal of NeuroEngineering and Rehabilitation

    (2011)
  • M. Guidali et al.

    A robotic system to train activities of daily living in a virtual environment

    Medical & Biological Engineering & Computing

    (2011)
  • Guidali, M., Schlink, P., Duschau-Wicke, A., & Riener, R. (2011). Online learning and adaptation of patient support...
  • Cited by (0)

    View full text