Modeling, Optimization, and Control Methods for a Personalized Hybrid Walking Exoskeleton

Overview

The central objective of this study is to validate new algorithms that coordinate between functional electrical stimulation (FES) and the exoskeleton during sitting-to-standing, walking, and standing-to-sitting movements. The secondary objective is to optimize the algorithms as well as assess their ability to reduce FES-induced muscle fatigue by using ultrasound imaging as a sensing modality. This study will include persons with no disabilities and persons with Spinal Cord Injury (SCI). A research set-up comprising of a lower-limb exoskeleton and FES system will be used to achieve sitting-to-standing, walking, and standing-to-sitting movements. Ultrasound Imaging probes may be used to record muscle activity of the stimulated muscles. The signals derived from ultrasound will be used to optimize FES in order to reduce muscle fatigue as well as assess muscle fatigue.

Research activities involve use of computer-controlled functional electrical stimulation (FES) and electric motor-driven orthosis (powered exoskeleton) to generate functional lower-limb movements for; i.e., standing and walking. The computer-controlled algorithms will get feedback from sensors such as optical encoders, load cells, force sensitive resistors, and inertial measurement units that are or will be inbuilt in the orthosis or these feedback algorithms will be combined with precomputed optimized algorithms based on the estimated musculoskeletal model of a participant. Multiple computer algorithms will be designed to coordinate FES with electric motors and other components of the walking device to reproduce or restore standing and walking movements. The study involves validation of computer algorithms to estimate and control the walking and sitting/standing movements. The Rifton E-Pacer motorized walker, arm crutches, parallel bars, or a conventional walker may be used in order to assist donning and doffing of the exoskeleton system, standing, and walking for all subjects, at any time during experimentation, for as long as is deemed necessary by the experimental goals. Using the E-Pacer or a walker should improve experiment safety and reduce the subject's overall physical exertion because it reduces load on the subject's body. Sitting/Standing and/or walking movements will be elicited by the hybrid walking platform that combines a powered exoskeleton and an FES system. The powered exoskeleton can provide joint actuation at the hip and knee joints of a participant. The FES system can stimulate the quadriceps, hamstrings muscle, glutes, and ankle muscles. The electrodes for stimulation are placed on the skin at locations that provide an optimal limb movement. The algorithms are designed such that the stimulation of different muscles are coordinated to achieve walking and/or sitting/standing. A precedent experiment will be performed on a Group B participant before it is tested on Group A. Unlike the paraplegic participants (Group A), if the device should fail or act unexpectedly the persons without disability (Group B) will be able to intervene . Therefore, testing the device on persons without disability first will allow us to further assess and address any risks before testing on individuals with SCI, who are incapable of intervening with the device. All the sessions for Group A participants will be performed in the supervision of a physical therapist. The therapist will also assist the Group A participants in transferring them from their wheelchairs to a cushioned table for seating if required or to the exoskeleton. Group A participant will also be supported by a body-unweighing system to prevent falling during standing and walking. Before each experiment involving electrical stimulation, a participant's heart rate and blood pressure will be recorded. Therapists will also assist the participants in transferring them from their wheelchairs to the therapy table. A participant or an individual operating the device will initiate walking and/or standing movement. A participant can stop the movements if he/she feels not comfortable or unsafe by using a safety-stop button, which will be given to them. During these sessions participants will wear a heart rate monitor to ensure that they do not overexert themselves, and so that sufficient rest periods may be given. The participants' heart rate will be checked before and after each trial by an individual conducting the experiments. The Borg scale will be administered for Group A to check for their exertion level. The sessions will be stopped as per the limits prescribed by the Borg scale (see stopping criteria in risks section). Please note that for all experiments described, a minimum of 1-2 days is required between visits for experimentation. These activities are divided into Aim 1 and Aim 2. Please note that Aims 1-2 are not sequential. Aim 1: To elicit sitting/standing and walking movements by using an optimized powered exoskeleton and functional electrical stimulation (FES) assistance. The objective of this aim is to experimentally validate an optimal controller that allocates FES and exoskeleton contribution based on a person-specific model of a participant. The controller will produce lower-limb movements in persons with no disabilities and persons with spinal cord injury. Group A and Group B will participate in approximately 20 sessions. The number of sessions is approximate and it may exceed this number and there is not cut-off number of sessions that will prohibit further participation. The session duration will be up to 4 hours, at the maximum. The session durations are approximate and it can be more or less than the specified. The total number sessions may overshoot by few sessions for some participants due to unanticipated technical or erroneous experimental data. These experiments will be conducted in the Neuromuscular Control and Robotics Laboratory (NCRL) or at the Outpatient Rehabilitation Clinic at Chapel-Hill run by the principal investigator. These experiments are divided into 2 tasks. Task 1: Initial visits for familiarization, training and model identification. Familiarization and training visits will allow us to customize and optimize the hybrid walking exoskeleton to each participating individual. For this task the participants will visit the NCRL for approximately 3 sessions. These sessions will take no more than 4 hours. The session durations are approximate and it can be more or less than the specified. The total number sessions may overshoot by few sessions for some participants due to unanticipated technical or erroneous experimental data. In the first visit, adjustments to the hybrid device will be made to fit the device to the participant. In these visits, the objectives are to train them for sit-to-stand transfers and perform walking movements with the hybrid walking device. Computer controlled algorithms will be used for performing sit-to-stand, stand-to-sit, and walking movements. Model identification experiments to find model parameters for each individual can be run during initial training and familiarization studies. These model identification experiments can be conducted while a participant sits/stands or walk in the hybrid walking device. Once the subject-specific parameters are identified, a walking and/or standing algorithm that coordinates FES with the powered exoskeleton will be derived or optimized using offline dynamic optimizations. Task 2: Validation of a sitting/standing and/or walking algorithm. After initial visits (Task 1), walking and/or standing algorithm will be tested. The algorithm is a computer program that takes sensory information from different sensors such optical encoders, inertial measurement units, and heel contact sensors in-built in the hybrid walking system to optimally coordinate FES and electric motors of the powered exoskeleton. The algorithm will be experimentally validated on each participant in approximately 7 sessions. Each session will run for no more than 4 hours, which includes time to perform standing, walking, periods of rest, time required to adjust bracing, and time required to update the computer code if needed. Note that the number of sessions and duration are estimates and are subject to change based on conditions of the experiment. A motion capture system may also be used to measure joint angles so that spatiotemporal gait characteristics can be compared with the normal gait of a healthy control subject. The recorded gait data will also be used to assess the performance of the control systems of the device, and to show how the user's gait changes over time as they become more comfortable with using the device. These sessions may be recorded so that the application and efficacy of the device may be illustrated. Aim 2: Integrating ultrasound-based feedback to control a hybrid exoskeleton. The objectives of this aim are to investigate if ultrasound imaging can be used as a tool to measure muscle fatigue induced by FES and further use the US-derived signals to optimize FES and exoskeleton assistance. The first objective of this aim is to develop an ultrasound based fatigue model to indicate the onset of muscle fatigue. The second objective is to use the ultrasound-based fatigue model in an algorithm that coordinates control of FES and electric motor during sitting-to-standing, walking, and standing-to-sitting activities. Group A and Group B individuals will participate in this aim. They will participate in approximately 10 sessions of no more than 4 hours each. The total number sessions can be more or less than 10 due to unanticipated technical or erroneous experimental data or if the participant leaves the study. These experiments will be conducted in NCRL or at the Outpatient Rehabilitation Clinic at Chapel-Hill. Task 1: Development of an ultrasound imaging based fatigue and recovery measures of leg muscles. A stimulation protocol will be followed that produces muscle fatigue. Muscle fatigue is defined as some percentage decline in isometric torque generated by FES of the quadriceps muscle. Because decline in the isometric torque could differ in different participants a common stimulation protocol will be employed for all the participants. This will be done during approximately two sessions that will each last no more than 4 hours. The session durations are approximate and it can be more or less than the specified. The total number sessions may overshoot by few sessions for some participants due to unanticipated technical or erroneous experimental data. Both ultrasound (using the imaging probe placed in between the electrodes) and torque data (from the load sensor) will be collected. A hand-held commercial ultrasound transducer (5- 20 MHz clinical transducer) connected to a commercial ultrasound system (S-Sharp Corporation, Taiwan) placed against the skin. A warm gel is applied to the skin to help with imaging. All the ultrasound imaging parameters including intensity and frame rates etc. are below safety guidelines for clinical ultrasound imaging (MI values less than 1.9). After these experiments, the processed ultrasound data will be correlated with the measured torque to determine fatigue and recovery measures. Task 2: Incorporation of the ultrasound-based fatigue model in a control allocation algorithm. The ultrasound imaging-based fatigue and recovery measures can be used to predict the onset of muscle fatigue. The experimental studies will be conducted on (Group A) and (Group B) over 2 sessions will each last no more than 4 hours. The session durations are approximate and it can be more or less than the specified. The total number sessions may overshoot by few sessions for some participants due to unanticipated technical or erroneous experimental data. The algorithms in Task 2 of Aim 1 will be modified to incorporate the ultrasound-based fatigue and recovery measures. Remark on stimulation parameters to be used in the aforementioned aims: A normal range of stimulation to be used in our studies is: pulse width 100-400 micro seconds; Frequency: 20-100 Hz; Current 0-100 mA. Generally during the experiment, the current is automatically modulated (or in some instances manually) starting from 0 mA to the value that is required to achieve a task. This is achieved by feedback signals to the control system that communicates with the stimulator. The control system does not allow the current to go above its maximum limit. The pulsewidth and frequency are generally kept constant. The stimulation parameters used for each participant will be documented for each experimental session and any changes in the stimulation parameters during a session will also be recorded.