Background: The limitations of functional electrical stimulation (FES) cycling directly affect the health benefits acquired from this technology and prevents its’ full potential to be realised. Experiments should be done on a test bed which can isolate and focus only on one muscle group, namely the quadriceps. The aim of this work was to design and develop an isokinetic robotic leg extension/flexion dynamometer which can mimic knee joint motion during actual cycling to be used for evaluation of novel functional electrical stimulation strategies. Although the main motivation for development of the dynamometer was for application in FES studies, it has the potential to be used for various different muscle physiology studies. Methods: A feedback control system with integrated electrical stimulation for isokinetic knee joint torque measurement has been developed and tested for safety and functionality. The leg extension/flexion device was modified and equipped with a DC motor drive system to imitate isokinetic knee joint motion during cycling when the hip joint remains fixed. Real-time bi-directional effective torque on the lever arm was measured by a magnetostrictive torque sensor and a load cell. Closed-loop motor control system was also designed to mimic the cyclical motion at desired angular velocity. Results: A functional model of the robotic dynamometer was developed and evaluated. The dynamometer is capable of simulating the knee angle during cycling at a cadence of up to 70 rpm with range of motion of 72 ◦. The magnetostrictive torque sensor can measure torque values up to 75 Nm. The lever arm can be adjusted and the target knee angle was controlled with RMSE tracking error of less than 2.1 ◦in tests with and without a test person, and with and without muscle stimulation. Conclusions: The isokinetic knee joint torque measurement system was designed and validated in this work, and subsequently used to develop and evaluate novel muscle activation strategies. This is important for fundamental research on effective stimulation patterns and novel activation strategies. This will, in turn, enhance the efficiency of FES cycling exercise and has the potential to improve the health-beneficial effects.

In order to promote early rehabilitation, we proposed a system which provides full-body arm-leg training for patients in a bed-lying position. As the preliminary development, a platform for leg movement was investigated. An innovative system with four servo drives was designed and manufactured. An artificial leg frame was attached to the platform via belts. The positions of the hip and knee joints were recorded using potentiometers. Closed-loop PID position control algorithms were implemented for production of various stepping movements. Technical evaluation on a test participant showed that the platform tracked the circular trajectory of the foot in a supine-lying position with an area difference of 8.2%, and produced walking-like trajectories in the hip and knee joints in a side-lying position with a mean error of 10.6%. The mechanical structure can be resized, and the control system can be expanded, so as to produce 3-dimensional stepping movement in both arms and legs. This innovative platform combined with the closed-loop position control strategy shows the technical potential to be a promising full-body rehabilitation platform for the patients in the early post-injury stage.

In a parallel development to traditional rigid rehabilitation robotic systems, cable-driven systems are becoming popular. The robowalk expander product uses passive elastic bands in the training of the lower limbs. However, a well-controlled assistance or resistance is desirable for effective walking relearning and muscle training. To achieve well-controlled force during locomotion training with the robowalk expander, we replaced the elastic bands with actuator-driven cables and implemented force control algorithms for regulation of cable tensions. The aim of this work was to develop an active cable-driven robotic system, and to evaluate force control strategies for walking rehabilitation using frequency-domain analysis. The system parameters were determined through experiment-assisted simulation. Then force-feedback lead controllers were developed for static force tracking, and velocity-feedforward lead compensators were implemented to reduce velocity-related disturbances during walking. The technical evaluation of the active cable-driven robotic system showed that force-feedback lead controllers produced satisfactory force tracking in the static tests with a mean error of 5.5%, but in the dynamic tests, a mean error of 13.2% was observed. Further implementation of the velocity-feedforward lead compensators reduced the force tracking error to 9% in dynamic tests. With the combined control algorithms, the active cable-driven robotic system produced constant force within the four cables during walking on the treadmill, with a mean force-tracking error of 10.3%. This study demonstrates that the force control algorithms are technically feasible. The active cable-driven, force-controlled robotic system has the potential to produce user-defined assistance or resistance in rehabilitation and fitness training.

Heart rate control using first- and second-order models was compared using a novel control design strategy which shapes the input sensitivity function. Ten participants performed two feedback control test series on a treadmill with square wave and constant references. Using a repeated measures, counterbalanced study design, each series compared controllers C1 and C2 based on first- and second-order models, respectively. In the first series, tracking accuracy root-mean-square tracking error (RMSE) was not significantly lower for C2: 2.59 bpm vs. 2.69 bpm (mean, C1 vs. C2), p = 0.79. But average control signal power was significantly higher for C2: 11.29 × 10^{−4} m2/s2 vs. 27.91 × 10^{−4} m2/s2, p = 3.1 × 10^{−10}. In the second series, RMSE was also not significantly lower for C2: 1.99 bpm vs. 1.94 bpm, p = 0.39; but average control signal power was again significantly higher for C2: 2.20 × 10^{−4} m2/s2 vs. 2.78 × 10^{−4} m2/s2, p = 0.045. The results provide no evidence that controllers based on second-order models lead to better tracking accuracy, despite the finding that they are significantly more dynamic. Further investigation using a substantially larger sample size is warranted.

Background: Accurate and robust feedback control of human heart rate is important for exercise testing and prescription. Feedback controllers can be designed using first-order, linear, time-invariant models of heart rate dynamics, but it remains to investigate whether second-order models lead to better identification and control performance. The distinguishing contribution of this research is the direct employment of established physiological principles to determine model structure, and to focus the feedbackdesign goals: cardiac physiology proposes a two-phase second-order response, delineated into fast and slow components; the natural phenomenon of broadspectrum heart-rate variability motivates a novel feedback design approach that appropriately shapes the input-sensitivity function. Aim: The aim of this work was to compare the fidelity of first- and second-order models of heart rate response during cycle-ergometer exercise, and to compare the accuracy and dynamics of feedback controllers that were designed using the two model structures. Methods: Twenty-seven participants each took part in two identification tests to generate separate estimation and validation data sets, where ergometer work rate was a pseudorandombinary sequence and in two feedback tests where controllers were designed using the first- or second-order models. Results: Second-order models gave substantially and significantly higher model fit (51.9 % vs. 47.9 %, p < 0.0001; second order vs. first order) and lower root-mean-square model error (2.93 bpm vs. 3.21 bpm, p < 0.0001). There was modest improvement in tracking accuracy with controllers based on second-order models, where mean root-mean-square tracking errors were 2.62 bpm (second order) and 2.77 bpm (first order), with p = 0.052. Controllers based on second-order models were found to be substantially and significantly more dynamic: mean values of average control signal power were 9.61 W^2 and 7.56 W^2, p < 0.0001. Conclusion: The results of this study confirm the hypotheses that second-order models of heart-rate dynamics give better fidelity than first-order models, and that feedback compensator designs that use the additional dynamic mode give more accurate and more dynamic closed-loop control performance.

The aim of this work was to investigate the time and exercise intensity dependence of heart rate variability (HRV). Time-dependent, cardiovascular-drift-related increases in heart rate (HR) were inhibited by enforcing a constant heart rate throughout the exercise with a feedback control system. Thirty-two healthy adults performed HR-stabilised treadmill running exercise at two distinct exercise intensity levels. Standard time and frequency domain HRV metrics were computed and served as outcomes. Significant decreases were detected in 8 of the 14 outcomes for the time dependence analysis and in 6 of the 7 outcomes for the exercise intensity dependence analysis (excluding the experimental speed-signal frequency analysis). Furthermore, metrics that have been reported to reach an intensity-dependent near-zero minimum rapidly (usually at moderate intensity) were found to be near constant over time and only barely decreased with intensity. Taken together, these results highlight that HRV generally decreases with time and with exercise intensity. The intensity-related reductions were found to be greater in value and significance compared to the time-related reductions. Additionally, the results indicate that decreases in HRV metrics with time or exercise intensity are only detectable as long as their metric-specific near-zero minimum has not yet been reached.

Aim and methods: The aim of this study was to compare the dynamics of heart rate (HR) response to exercise using a cycle ergometer (CE) and a treadmill (TM). Using a sample of 25 healthy male participants, the time constant of HR dynamics was estimated for both modalities in response to square-wave excitation. Results: The principal finding was that the time constant of heart-rate dynamics around somewhat hard exercise intensity (Borg rating of perceived exertion = 13) does not differ significantly between the CE and TM (68.7 s ± 21.5 s vs. 62.5 s ± 18.5 s [mean ± standard deviation]; CE vs. TM; p = 0.20). An observed moderate level of evidence that root-mean-square model error was higher for the CE than for the TM (2.5 bpm ± 0.5 bpm vs. 2.2 bpm ± 0.5 bpm, p = 0.059) may reflect a decrease in heart rate variability with increasing HR intensity because, in order to achieve similar levels of perceived intensity, mean heart rate for the CE was approximately 25 bpm lower than for the TM. Conclusion and significance: These results have important implications for model-based design of automatic HR controllers, because, in principle, the same dynamic controller, merely scaled according to the differing steady-state gains, should be able to be applied to the CE and TM exercise modalities.

Background: Characterisation of heart rate (HR) dynamics and their dependence on exercise intensity provides a basis for feedback design of automatic HR control systems. This work aimed to investigate whether the second-order models with separate Phase I and Phase II components of HR response can achieve better fitting performance compared to the first-order models that do not delineate the two phases. Methods: Eleven participants each performed two open-loop identification tests while running at moderate-to-vigorous intensity on a treadmill. Treadmill speed was changed as a pseudo-random binary sequence (PRBS) to excite both the Phase I and Phase II components. A counterbalanced cross-validation approach was implemented for model parameter estimation and validation. Results: Comparison of validation outcomes for 22 pairs of first- and second-order models showed that root-mean-square error (RMSE) was significantly lower and fit (normalised RMSE) significantly higher for the second-order models: RMSE was 2.07 bpm ± 0.36 bpm vs. 2.27 bpm ± 0.36 bpm (bpm = beats per min), second order vs. first order, with p = 2.8 × 10^{−10} ; fit was 54.5% ± 5.2 % vs. 50.2% ± 4.8 %, p = 6.8 × 10^{−10}. Conclusion: Second-order models give significantly better goodness-of-fit than firstorder models, likely due to the inclusion of both Phase I and Phase II components of heart rate response. Future work should investigate alternative parameterisations of the PRBS excitation, and whether feedback controllers calculated using second-order models give better performance than those based on first-order models.

Background: Interlimb neural coupling implies that arm swing should be included during gait training to improve rehabilitation outcomes. We previously developed several systems for production of walking with arm swing, but the reaction forces on the foot sole during usage of the systems were not satisfactory and there was potential to improve control system performance. This work aimed to design and technically evaluate a novel system for producing walking with synchronised arm and leg movement and with dynamic force loading on the foot soles. Methods: The robotic system included a passive curved treadmill and a trunk frame, upon which the rigs for the upper and lower limbs were mounted. Ten actuators and servocontrollers with EtherCAT communication protocol controlled the bilateral shoulder, elbow, hip, knee and ankle joints. Impedance control algorithms were developed and ran in an industrial PC. Flexible pressure sensors recorded the plantar forces on the foot soles. The criteria of implementation and responsiveness were used to formally evaluate the technical feasibility of the system. Results: Using impedance algorithms, the system produced synchronous walking with arm swing on the curved treadmill, with mean RMS angular tracking error <2° in the 10 joint profiles. The foot trajectories relative to the hip presented similar shapes to those during normal gait, with mean RMS displacement error <1.5 cm. A force pattern that started at the heel and finished at the forefoot was observed during walking using the system, which was similar to the pattern from overground walking. Conclusion: The robotic system produced walking-like kinematics in the 10 joints and in the foot trajectories. Integrated with the curved treadmill, the system also produced walking-like force patterns on the foot soles. The system is considered feasible as far as implementation and responsiveness are concerned. Future work will focus on improvement of the mechanical system for future clinical application.

BACKGROUND: To provide effective rehabilitation in the early post-injury stage, a novel robotic rehabilitation platform is proposed, which provides full-body arm-leg rehabilitation via belt actuation to severely disabled patients who are restricted to bed rest. OBJECTIVE: To design and technically evaluate the preliminary development of the rehabilitation platform, with a focus on the generation of various leg movements. METHODS: Two computer models were developed by importing the components from SolidWorks into Simscape Multibody in MATLAB. This allowed simulation of various stepping movements in supine-lying and side-lying positions. Two belt-actuated test rigs were manufactured and automatic control programs were developed in TIA Portal. Finally, the functionality of the test rigs was technically evaluated. RESULTS: Computer simulation yielded target positions for the generation of various stepping movements in the experimental platforms. The control system enabled the two-drive test rig to provide three modes of stepping in a supine position. In addition, the four-drive test rig produced walking-like stepping in a side-lying position. CONCLUSIONS: This work confirmed the feasibility of the mechanical development and control system of the test rigs, which are deemed applicable for further development of the overall novel robotic rehabilitation platform.

The objective was to assess the performance and robustness of a novel strategy for automatic control of heart rate (HR) during cycle ergometry. Control design used a linear plant model and direct shaping of the closed-loop input-sensitivity function to achieve an appropriate response to disturbances attributable to broad-spectrum heart rate variability (HRV). The controller was evaluated in 73 feedback control experiments involving 49 participants. Performance and stability robustness were analysed using a separately identified family of 73 plant models. The controller gave highly accurate and stable HR tracking performance with mean root-mean-square tracking error between 2.5 beats/min (bpm) and 3.1 bpm, and with low average control signal power. Although plant parameters varied over a very wide range, key closed-loop transfer functions remained invariant to plant uncertainty in important frequency bands, while infinite gain margins and large phase margins (>62◦) were preserved across the whole plant model family. Highly accurate, stable and robust HR control can be achieved using LTI controllers of remarkably simple structure. The results highlight that HR control design must focus on disturbances caused by HRV. The input-sensitivity approach evaluated in this work provides a transparent method of addressing this challenge.

Introduction: With conventional heart rate (HR) control systems, the exercising person is bound to walk or run at a pace determined by the feedback. This may be challenging for people with impairments that make it difficult for them to achieve a smooth, continuous pace. The aim of this work was to assess the technical feasibility of a novel self-paced heart rate control strategy and to compare its accuracy with conventional heart rate control. Methods: We propose a self-paced heart rate control system that embeds an automatic positioning controller within the heart rate control loop. The treadmill speed command is decoupled from the heart rate compensator, whereas speed is determined by the exerciser’s own volition: target speed is displayed visually to the person and, when they try to follow this target, the position controller sets the treadmill speed while keeping the person at a safe reference position on the track. A further novel contribution of this work is a new input-sensitivity-shaping, frequency-domain design strategy for feedback control of position. Results: Experimental evaluation with four participants showed that self-paced heart rate control is technically feasible: all participants were able to accurately follow the target running speed calculated by the HR compensator and presented to them visually; for all four participants, self-paced HR tracking accuracy was not substantially different from conventional HR control performance; on average, the self-paced heart rate controller gave slightly better performance than conventional HR control, with RMS tracking error of 2.98 beats per minute (bpm) vs 3.11 bpm and higher average control signal power. Conclusion: The proposed self-paced heart rate control strategy with embedded automatic position control is deemed feasible. This approach may be helpful for people with gait impairments or other limitations that make it difficult for them to follow an imposed treadmill speed.

Background: During functional electrical stimulation (FES) tasks with able-bodied (AB) participants, spatially distributed sequential stimulation (SDSS) has demonstrated substantial improvements in power output and fatigue properties compared to conventional single electrode stimulation (SES). The aim of this study was to compare the properties of SDSS and SES in participants with spinal cord injury (SCI) in a dynamic isokinetic knee extension task simulating knee movement during recumbent cycling. Method: Using a case-series design, m. vastus lateralis and medialis of four participants with motor and sensory complete SCI (AIS A) were stimulated for 6 min on both legs with both electrode setups. With SES, target muscles were stimulated by a pair of electrodes. In SDSS, the distal electrodes were replaced by four small electrodes giving the same overall stimulation frequency and having the same total surface area. Torque was measured during knee extension by a dynamometer at an angular velocity of 110 deg/s. Mean power of the left and right sides (PmeanL,R) was calculated from all stimulated extensions for initial, final and all extensions. Fatigue is presented as an index value with respect to initial power from 1 to 0, whereby 1 means no fatigue. Results: SDSS showed higher PmeanL,R values for all four participants for all extensions (increases of 132% in participant P1, 100% in P2, 36% in P3 and 18% in P4 compared to SES) and for the initial phase (increases of 84%, 59%, 66%, and 16%, respectively). Fatigue resistance was better with SDSS for P1, P2 and P4 but worse for P3 (0.47 vs 0.35, 0.63 vs 0.49, 0.90 vs 0.82 and 0.59 vs 0.77, respectively). Conclusion: Consistently higher PmeanL,R was observed for all four participants for initial and overall contractions using SDSS. This supports findings from previous studies with AB participants. Fatigue properties were better in three of the four participants. The lower fatigue resistance with SDSS in one participant may be explained by a very low muscle activation level in this case. Further investigation in a larger cohort is warranted.

Background: Leg-press devices are one of the most widely used training tools for musculoskeletal strengthening of the lower-limbs, and have demonstrated important cardiopulmonary benefits for healthy and patient populations. Further engineering development was done on a dynamic leg-press for work-rate estimation by integrating force and motion sensors, power calculation and a visual feedback system for volitional work-rate control. This study aimed to assess the feasibility of the enhanced dynamic leg press for cardiopulmonary exercise training in constant-load training and high-intensity interval training. Five healthy participants aged 31.0 ± 3.9 years (mean ± standard deviation) performed two cardiopulmonary training sessions: constant-load training and high-intensity interval training. Participants carried out the training sessions at a work rate that corresponds to their first ventilatory threshold for constant-load training, and their second ventilatory threshold for high-intensity interval training. Results: All participants tolerated both training protocols, and could complete the training sessions with no complications. Substantial cardiopulmonary responses were observed. The difference between mean oxygen uptake and target oxygen uptake was 0.07 ± 0.34 L/min (103 ±17%) during constant-load training, and 0.35 ± 0.66 L/min (113 ±27%) during high-intensity interval training. The difference between mean heart rate and target heart rate was −7 ± 19 bpm (94 ±15%) during constant-load training, and 4.2 ± 16 bpm (103 ±12%) during high-intensity interval training. Conclusions: The enhanced dynamic leg press was found to be feasible for cardiopulmonary exercise training, and for exercise prescription for different training programmes based on the ventilatory thresholds.

To investigate the time dependence of the heart rate variability (HRV) during treadmill running, a feedback control loop was implemented to eliminate the potentially confounding influence of cardiovascular drift. Without cardiovascular drift, observed changes in HRV can be directly attributed to time only and not to drift-related increases in heart rate. To quantify the time-dependence of HRV, standard HRV metrics for two consecutive windows of equal duration (12.5 min) were computed and compared. Eight participants were included. The outcome measures showed an overall tendency to decrease over time. Seven of the 10 HRV metrics were significantly lower (p<0.05); three HRV metrics showed moderate evidence of decrease over time, viz. average control power P∇u (p = 0.053), very-low frequency power (VLF) of the RR-signal (p = 0.072) and low frequency power (LF) of the RR-signal (p = 0.12). Taken together, these results provide evidence of a decrease in HRV over time during treadmill running; the employment of feedback control of heart rate is important as cardiovascular drift was eliminated. Further work is required to optimize the experimental design and to use a larger sample size to improve the statistical power of the results.

Objective: To develop a unified heart rate (HR) control approach for cycle ergometer (CE) and treadmill(TM) exercise, and to empirically compare the common controller’s performance between the CE andTM. Methods: The control method used frequency-domain shaping of the input-sensitivity function to addressrejection of disturbances arising from broad-spectrum heart rate variability (HRV). A single controllerwas calculated using an approximate, nominal linear plant model and an input-sensitivity bandwidthspecification. Fifty HR control tests were executed using the single controller: 25 healthy male participantseach did one test on the CE and one on the TM. Results: There was no significant difference in mean root-mean-square HR tracking error: 3.10 bpm ±0.68 bpm and 2.85 bpm ± 0.75 bpm (mean ± standard deviation, bpm = beats/min); CE vs. TM; p = 0.13.But mean normalised average control signal power was significantly different: 1.59 bpm2± 0.27 bpm2vs. 1.36 bpm2± 0.28 bpm2; CE vs. TM; p = 3.5 × 10−4. Conclusion and significance: The lower values for RMS tracking error and control signal power for the TMpoint to decreasing HRV intensity with increasing HR, because, in order to match perceived exertion forthe two modalities, mean HR for the TM was set 20 bpm higher than for the CE. These HR-intensity-dependent differences in HRV are consistent with previous observations in the literature. The unified HRcontrol approach for CE and TM exercise gave accurate, stable and robust performance in all tests, thuslending support to the concept that HRV disturbance rejection is the main issue in HR control design.

BACKGROUND: The use of robotic technology for neurorehabilitative applications has become increasingly important for adults and children with different motor impairments. OBJECTIVE: The aim of this study was to evaluate the technical feasibility and usability of a new interactive leg-press training robot that was developed to train leg muscle strength and control, suitable for children with neuromuscular impairments. METHODS: An interactive robotic training system was designed and constructed with various control strategies, actuators and force/position sensors to enable the performance of different training modes (passive, active resistance, and exergames). Five paediatric patients, aged between 7 and 16 years (one girl, age 13.0 +/- 3.7 years, [mean +/- SD]), with different neuromuscular impairments were recruited to participate in this study. Patients evaluated the device based on a user satisfaction questionnaire and Visual Analog Scale (VAS) scores, and therapists evaluated the device with the modified System Usability Scale (SUS). RESULTS: One patient could not perform the training session because of his small knee range of motion. Visual Analog Scale scores were given by the 4 patients who performed the training sessions. All the patients adjudged the training with the interactive device as satisfactory. The average SUS score given by the therapists was 61.2 +/- 18.4. CONCLUSION: This study proposed an interactive lower limb training device for children with different neuromuscular impairments. The device is deemed feasible for paediatric rehabilitation applications, both in terms of technical feasibility and usability acceptance. Both patients and therapists provided positive feedback regarding the training with the device.