Feasibility of Hybrid Gait Training with Kinesis Overground Robot for

Persons with incomplete Spinal Cord Injury

Antonio J. del-Ama

1,2

Angel Gil-Agudo

, Jos

e L. Pons

and Juan C. Moreno

Biomechanics and Technical Aids Department, National Hospital for Spinal Cord Injury, Toledo, Spain

Bioengineering Group of the Spanish National Research Council, Madrid, Spain

Keywords:

Spinal Cord Injury, FES, Exoskeleton, Gait Training.

Abstract:

Hybrid actuation and control have a considerable potential for walking rehabilitation but there is a need of

novel hybrid control strategies that adequately manage the balance between FES and robotic controllers. A

hybrid co-operative control strategy for overground gait training with a wearable robotic exoskeleton for per-

sons with incomplete spinal cord injury (SCI) is presented. The feasibility of the control strategy to overcome

muscular stimulation electro-mechanical delay, deterioration of muscle performance over time, and to balance

muscular and robotic actuation cyclic overground walking is tested in one subject with incomplete spinal cord

injury (L4, ASIA grade D). The results demonstrate that the proposed hybrid cooperative control in Kinesis

overground robot is able to autonomously compensate a bilateral pathologic walking pattern and the suitability

of Kinesis hybrid gait training robot for conducting clinical experimentation.

1 BACKGROUND

Most therapies for rehabilitation of walking rely on

the assumption that task-oriented practice promotes

mechanisms of neural plasticity, muscle strength

and learning of compensation strategies that increase

walking ability of the person with SCI. Robotic tech-

nology holds a considerable potential to drive such

interventions. Ambulatory robots, that have been de-

veloped mainly for functional compensation of walk-

ing, can offer a challenging and rich walking ther-

apy. Furthermore, functional electrical stimulation

can drive rehabilitation interventions of SCI provid-

ing several physiological and psychological beneﬁts

to the user through artiﬁcial activation of paralyzed

muscles. On the other hand, robotic exoskeletons can

be used to manage the unavoidable loss of perfor-

mance of FES-driven muscles. Hybrid exoskeletons

are then regarded as a promising approach that blends

complementary robotic and neuroprosthetic technolo-

gies. The overview of the state of the art on hybrid

gait systems has demonstrated that such redundant

actuated solutions can produce feasible systems for

accurate control of joint movement (Del-Ama et al.,

2012a). Also, it has shown that diverse muscle fa-

tigue management strategies could be applied for an

effective closed-loop control of FES.

Under this hybrid scenario, the assist-as-needed

control strategy has been proposed as a new redun-

dant neuroprosthetic and robotic system that cooper-

ates to optimize the outcome of the active control of

the motion of the knee joint while providing assis-

tance (Del-Ama et al., 2013). While various wearable

exoskeletons were successful in achieving gait in sub-

jects with incomplete SCI (Dollar and Herr, 2007),

(Dollar A.M.; Herr, 2008), (Hesse et al., 2010) this

has generally been proposed as a functional substitu-

tion. In this case study, we tested the feasibility of

a novel hybrid gait training approach with an over-

ground robot on a person with an incomplete SCI.

2 MATERIAL AND METHODS

2.1 Subject

Target population of Kinesis is comprised by patients

whose SCI is located under the second lumbar ver-

tebrae. The subject was a 43-year old male (75 kg.

and 1.77 m. height). The subject suffered a trau-

matic lesion at L4 (ASIA grade D). The subject met

the criteria of the target population of Kinesis, cate-

gorized as Conus Medularis (Hayes et al., 2000). The

prognosis of functional recovery of walking is that the

subject could walk short distances but depending on

190

del-Ama A., Gil-Agudo Á., Pons J. and Moreno J..

Feasibility of Hybrid Gait Training with Kinesis Overground Robot for Persons with incomplete Spinal Cord Injury.

DOI: 10.5220/0004659701900193

In Proceedings of the International Congress on Neurotechnology, Electronics and Informatics (RoboAssist-2013), pages 190-193

ISBN: 978-989-8565-80-8

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

the wheelchair for community ambulation, therefore a

successful hybrid walking therapy could provide ben-

eﬁts to this population. The subject has a preserved

hip ﬂexion ability, partial ability to generate volun-

tary knee extension and is in presence of mild spastic-

ity. The subject provided written informed consent by

signing a form that was approved by the Spinal Cord

National Hospital Review Board. As consequence of

the accident, the patient had a limited articular range

at both knees, which led to adaptation of the kine-

matic pattern of the left leg to meet these physical

constraints. The modiﬁcation consisted on increasing

the stance angle to met the maximum extension an-

gle for the left leg, and to scale the ﬂexion-extension

pattern from stance angle to 60 degrees.

2.2 Protocol

The subject participated in the hybrid gait training

testing session (HGTT) to determine the feasibility of

overground control of walking with Kinesis. Prior to

the HGTT, the patient underwent a stimulation test

and a training session. The stimulation test was em-

ployed to quantify the muscular response to the mus-

cle stimulation and also to get the patient used to the

stimulation. Within this stimulation test, both ﬂexor

and extensor knee muscle groups of both legs were

stimulated for 15 minutes. Then, a training session

took place in which the subject carried out learning

exercises with the Kinesis system. In this training ses-

sion the basic walking technique was explained to the

user (bend to the side to lift the heel prior to initiate a

step and then pressing a manual button). Kinesis was

adjusted to the patient anthropometry within this ses-

sion. Total time walking in this training session did

not exceed from 10 minutes. During HGTT kinesis

hybrid-cooperative controller modulated both stimu-

lation and robotic assistance during walking.

2.3 Hybrid Cooperative Control in

Kinesis

Kinesis is a hybrid robotic device that has been devel-

oped for overground gait training in incomplete spinal

cord injuries. The Kinesis system is a bilateral wear-

able knee-ankle-foot orthosis, equipped with active

actuators at the knee hinges (Maxon DC ﬂat motor,

90W with Harmonic-Drive 100:1 gear) and a passive

elastic actuator at the ankle. Force sensing resistors

are employed for monitoring ﬂoor contact and cus-

tom force sensors are available to measure interac-

tion torques. Kinesis has a PC-controlled stimulator

(Rehastim, Hasomed GmbH) which delivers biphasic

current-controlled rectangular pulses. Rehastim can

be pulse width and current controlled in real time.

The high-level control approach to achieve a coop-

erative behavior comprises four main components: 1)

robotic or joint controller, 2) FES controller, 3) mus-

cle fatigue estimator (MFE), and 4) a ﬁnite-state ma-

chine (FSM) that orchestrates the FES and joint con-

trollers. Further details on the implementation of the

high-level control can be found in (Del-Ama et al.,

2012b) and (Del-Ama et al., 2013). The cooperative

behavior of Kinesis allows obtaining adequate and

personalized stimulation patterns, estimating muscle

fatigue and reducing robotic assistance during ambu-

latory walking. The ultimate goal is to give priority to

the muscle-generated torque during gait training. A

ﬁnite-state machine is employed to iteratively control

the FES of knee muscles in a learning scheme dur-

ing for each leg while adapting torque ﬁeld stiffness

for a reference kinematic pattern. In this scheme, the

resulting interaction torque (with added mass and in-

ertia of the leg) is monitored and used towards con-

vergence of stimulation parameters. The robot mod-

ulates its assistance by reducing joint stiffness and

ensuring the target ﬂexion angle for effective swing

of the leg. A muscle fatigue estimator is employed

(based on the measurement of interaction torque) to

trigger a fatigue compensation strategy (change stim-

ulation ﬁring rate). More detailed descriptions of the

technique for hybrid cooperative control of Kinesis

are discussed in (Del-Ama et al., 2013) and (Del-Ama

et al., 2012b).

2.4 Robot Stiffness Modulation Strategy

The strategy to modulate the exoskeletal knee stiff-

ness during cyclic walking is described in this sec-

tion. The efﬁcacy of the FES controller to gener-

ate the knee movement is inherently limited, due to

the low efﬁciency of the force generated by the stim-

ulated muscles and the electromechanical delay be-

tween the stimulus and the onset of joint movement.

The goal of the hybrid control strategy was therefore

to exploit the joint movement generated by the NP

while supporting the movement through the NR. A

controller was employed to provide compliant assis-

tance to the knee, depending upon the parameter K

the stiffness of the force ﬁeld applied around the tra-

jectory. Modulation of K

was executed depending

on the gait phase and the contribution of the FES to

the knee trajectory. Thus, gait phase and muscle con-

tribution were managed within a ﬁnite state machine

(FSM), comprised of two FSMs operating in parallel:

one FSM runs in the time domain (t-FSM) while the

other operates in the cycle domain (c-FSM). The t-

FSM detected the main walking states and the transi-

FeasibilityofHybridGaitTrainingwithKinesisOvergroundRobotforPersonswithincompleteSpinalCordInjury

191

tions among them. The c-FSM operated with discrete

values, during each swing phase, that are related to

performance of the stimulated muscle. This allows to

uncouple the closed-loop control of stimulation from

the muscle fatigue monitoring and management.

Muscle fatigue results in a decrease of muscle per-

formance thus increasing the interaction torque. This

increase can be automatically compensated with the

closed-loop FES controller reducing the interaction

torque.

The t-FSM modulated the force ﬁeld stiffness K

and set the kinematic pattern, depending on the walk-

ing state. The compliant behavior of the exoskeleton

was achieved by controlling knee trajectory through a

ﬁrst order torque ﬁeld imposed around the joint trajec-

tory. As a result, the joint torque imposed by the robot

depended on the deviation of the knee trajectory from

the kinematic pattern and the stiffness of the torque

ﬁeld K

. The width of a virtual tunnel where the knee

can actually move could be adjusted along time. Dur-

ing stance, a high stiffness torque ﬁeld was imposed

to provide sufﬁcient support and avoid knee collapse.

Conversely, the supportive actions of the exoskeleton

must be reduced during the swing phase to allow for

the contribution of stimulated muscles and passive dy-

namics. The former requirement was achieved by re-

ducing the support of the robot through a wider vir-

tual tunnel. At the end of each swing phase, prior to

contacting the ﬂoor, the robot gradually increased its

support to ensure full knee extension, through a pro-

gressive increment in the force ﬁeld stiffness. How-

ever, this late stiffness for foot contact is insufﬁcient

for weight support and therefore, a quick transition to

high stiffness required for stance was implemented.

2.5 Biomechanical Evaluation

The analysis of feasibility was performed at the

biomechanical level. We assessed the actual knee

joint kinematics and stiffness during overground hy-

brid control of gait.

3 RESULTS

We describe the observed performance of the over-

ground hybrid control during the ﬁrst and last steps of

the HGTT followed by the observed evolution in the

automatic control of the stimulation of the quadriceps

muscles.

Figure 1: Results of HGTT session (cycle domain) for left

(top) and right (bottom) legs. Controller stiffness (magenta

curve), normalized torque-time integral (NTTI, red curve),

maximum angle achieved during ﬂexion (blue curve) and

normalized stimulation intensity for knee ﬂexor muscles

(NILC, black curve) of both legs. Active learning state

is represented by green boxes, otherwise monitoring state.

Controller stiffness, maximum angle and normalized stim-

ulation curves are scaled.

3.1 Stiffness Modulation Performance

and FES Management

The stimulation pattern of both legs is represented in

the cycle-domain (Figure 1). The NILC of the right

leg reached a 60% of the maximum stimulation in-

tensity achievable for the swing phase in the ﬁrst it-

eration, thus reacting to the voluntary knee extension

exerted by the patient. The active contribution of the

patient had also an impact on the interaction forces, as

shown in the Figure 1, in particular in the progressive

increase on NTTI (normalized time torque integral)

values along the ﬁrst 7 steps.

These can be explained as a particular strategy

adopted by the patient towards the robotic and neu-

roprosthetic assistance. After the ﬁrst swing cycle,

the patient reacted by increasing the active knee ex-

tension during consecutive cycles, and thus, increas-

ing the NTTI. It is noticeable that after the learning

period (cycle 14), the stiffness was automatically in-

creased as a response to the augmented active exten-

sion, targeting 60 degrees of knee ﬂexion (Figure 1b).

After the ﬁrst learning period, a decrease on NTTI

is observed, explained by the stimulation effect and

NEUROTECHNIX2013-InternationalCongressonNeurotechnology,ElectronicsandInformatics

192

patient learning. Then, until cycle 27, fatigue was-

detected and a new learning period took place. Con-

versely, the increase on the stimulation intensity for

the left leg muscle was signiﬁcantly slow. After the

learning period was completed for this muscle, the

stiffness could not be signiﬁcantly reduced to accom-

plish the kinematic pattern.

Figure 2: Normalized stimulation of the right and left

quadriceps muscles during stance with hybrid cooperative

control (x axis: number of steps).

Figure 2 depicts the averaged quadriceps stimula-

tion intensity for the stance phases for both legs dur-

ing the HGTT. It is noticeable the high stimulation in-

tensity applied to the left leg, in response to a ﬂexion

posture exhibited by the patient during stance phases.

Furthermore, as the experiment progressed and more

steps were taken, the stimulation intensity progres-

sively augmented until reaching a plateau (step 13).

4 DISCUSSION

The speciﬁc functional deﬁcit of the subject lead to

a limitation on the maximum (left) knee extension,

which caused the patient to exert compensating ac-

tions. These compensations were effectively counter-

acted by the hybrid gait control. The compensating

actions differed for both side and stance and swing

phases of gait. During stance, on average, the pa-

tient ﬂexed the right knee in an attempt to compen-

sate for the ﬂexion angle of the left knee. The robotic

actuation compensated the stimulation as needed: the

displayed stiffness was sufﬁcient to provide compli-

ant but adequate support during stance phases. We

also noticed that during pre-swing phases, the subject

consistently changed from ﬂexion to extension, prob-

ably as a response to the limited range of movement

of the right knee. We concluded that the hybrid co-

operative control in Kinesis is able to compensate a

bilateral pathologic walking pattern by autonomously

increasing the stimulation of the ﬂexor muscles and

increasing the displayed stiffness of the robotic actu-

ator.

REFERENCES

Del-Ama, A. J., Gil-Agudo, A., Pons, J. L., and Moreno,

J. C. (2013). Hybrid FES-robot cooperative control of

ambulatory gait rehabilitation exoskeleton for spinal

cord injured users. Manuscript submitted for publica-

tion (Mar 2013).

Del-Ama, A. J., Koutsou, A. D., and Moreno, J. C. (2012a).

Review of hybrid exoskeletons to restore gait follow-

ing spinal cord injury. Journal of Rehabilitation Re-

search and Development, 49(4):497–514.

Del-Ama, A. J., Moreno, J. C., Gil-agudo, A., and Pons,

J. L. (2012b). Hybrid FES-robot cooperative con-

trol of ambulatory gait rehabilitation exoskeleton for

spinal cord injured users. In Pons, J. L., Torricelli, D.,

and Pajaro, M., editors, 2012 International Confer-

ence on Neurorehabilitation (ICNR2012): Converg-

ing Clinical and Engineering Research on Neuroreha-

bilitation, Toledo (Spain). Springer Heidelberg New

York.

Dollar, A. M. and Herr, H. (2007). Active Orthoses for the

Lower-Limbs: Challenges and State of the Art. IEEE

International Conference on Rehabilitation Robotics,

1:968–977.

Dollar A.M.; Herr, H. (2008). Lower Extremity Exoskele-

tons and Active Orthoses: Challenges and State-of-

the-Art. IEEE Transactions on Robotics, 24(1):144–

158.

Hayes, K., Hsieh, J., Wolfe, D., Potter, P., and Delaney, G.

(2000). Classifying incomplete spinal cord injury syn-

dromes: Algorithms based on the international stan-

dards for neurological and functional classiﬁcation

of spinal cord injury patients. Archives of physical

medicine and rehabilitation, 81(5):644–652.

Hesse, S., Waldner, A., and Tomelleri, C. (2010). Inno-

vative gait robot for the repetitive practice of ﬂoor

walking and stair climbing up and down in stroke pa-

tients. Journal of NeuroEngineering and Rehabilita-

tion, 7:30.

FeasibilityofHybridGaitTrainingwithKinesisOvergroundRobotforPersonswithincompleteSpinalCordInjury

193