Robust control of heart rate for cycle ergometer exercise
Abstract
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 closedloop inputsensitivity function to achieve an appropriate response to disturbances attributable to broadspectrum 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 rootmeansquare 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 closedloop 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 inputsensitivity approach evaluated in this work provides a transparent method of addressing this challenge.
Keywords
Heart rate control Heart rate dynamics Heart rate variability System identification Physiological control Cycle ergometersGlossary of terms
 BMI:
Body mass index
 bpm:
Beats/min
 HR:
Heart rate
 HR^{∗}:
Target heart rate
 HRmax:
Maximal heart rate
 HRmid:
Midlevel heart rate
 HRnom:
Nominal (simulated) heart rate
 HRMI:
Heart rate monitor interface
 HRV:
Heart rate variability
 k:
Steadystate gain
 LTI:
Linear, timeinvariant
 P_{∇u}:
Average control signal power
 PC:
Personal computer
 PI:
Proportionalintegral
 PID:
Proportionalintegralderivative
 RMS:
Rootmeansquare
 RMSE:
RMS tracking error
 RPE:
Rating of perceived exertion
 rpm:
Revolutions per minute
 τ:
Time constant
 USB:
Universal serial bus
 v:
Speed
 WR:
Work rate
1 Introduction
Wellestablished guidelines exist for the definition of cardiopulmonary exercise testing protocols and for the prescription of training regimes. Specific testing and prescription guidelines are available for healthy individuals and for patients across a diversity of health conditions [1]; the most common exercise modalities are treadmill walking/running and cycle ergometry, while exercise intensity can be characterised using such variables as heart rate (HR), oxygen uptake or a subjective rating of perceived exertion, RPE [2].
HR is a quantitative variable that can be easily measured, and several approaches have been investigated for automatic control of HR during both treadmill and cycle ergometer exercise ([3, 4, 5] and [6, 7, 8], respectively). These feedback systems facilitate tracking of arbitrary HR profiles by automatically and continuously adjusting a manipulated variable, which for treadmills can be speed or slope, or both, and which for cycling is usually work rate.
Since cycle ergometers provide a stableseated position, they are the preferred modality for exercise testing and prescription in cardiac rehabilitation; HR controllers have long been investigated in this context [6, 9, 10], but also, subsequently, for healthy persons [7, 8, 11].
A most elegant treatment of HR control for cycle ergometers was provided by Kawada et al. [6]. In that work, a single linear transfer function model of HR response to changes in work rate was obtained as an average from openloop system identification experiments with 10 individual participants (8 men, 2 women). The model was then used in simulation to tune the two free parameters of a linear proportionalintegral (PI) controller. The single timeinvariant PI controller thus obtained was then tested in HR control experiments with 55 healthy participants (45 men, 10 women) and with 12 patients with cardiac disease (10 men, 2 women). In the healthy participants (n = 55), the mean rootmeansquare tracking error (RMSE) for a constant HR target of 60 % of maximal HR (HRmax) was 2.5 beats/min (bpm); when the HR target was 75 % of HRmax, mean RMSE was 3.8 bpm. For the cardiac patients (n = 12) exercising at a constant target HR of 20 bpm above resting HR, mean RMSE was 3.0 bpm. This work, which reported HR control data from 122 individual HR control tests with 67 participants in the two experimental cohorts, thus provides strong empirical evidence that a single linear, timeinvariant (LTI) controller of very simple structure can provide accurate and robust HR control.
A variety of nonlinear approaches to HR modelling and control for both treadmills and cycle ergometers have been proposed. Nonlinear models have been used to represent the different gains and time constants that exist for positive and negative step changes in speed [12]; asymmetry has also been observed and modelled during moderateintensity treadmill running [13]. For the purpose of control design, a nonlinear statespace model, where the control signal appears in quadratic form, was employed and combined with linearquadratic and Hinfinity optimisation [14]; the same model structure was used, but with a nonlinearitycancellation strategy, for HR control using a treadmill [4] or cycle ergometer [7]. A related approach using a Hammerstein model structure and a compensator with cancellation of the nonlinear model term was combined with modelpredictive control [15]. Other approaches include linear Hinfinity control with static nonlinearity compensation [16] and a nonlinear neural network approach [17]. A limitation common to most of these reports is that quantitative measures of controller performance (i.e. RMSE and control signal intensity) were not employed and that very small numbers of participants were included in experimental evaluations, thus making it difficult to objectively gauge their utility.
A recent study of HR control during cycle ergometry combined an LTI proportionalintegralderivative (PID) controller with an auditory biofeedback signal [8]. Despite the humanintheloop nature of this approach, quite accurate tracking was achieved with mean RMSE on the range 3.7 bpm to 5.0 bpm (various experiments with 24 healthy male participants).
In concordance with some of the above observations, a growing body of evidence has emerged from treadmill studies that points towards heart rate variability (HRV, [18]) as the principal challenge in the design of HR control systems, in contradistinction to parametric and/or structural sources of plant uncertainty. From a controltheoretical perspective, HRV presents as a broadspectrum disturbance signal [19]; care must therefore be taken to ensure that the control signal is not unduly excited at frequencies that might disturb the exercising subject. In short, the said studies have demonstrated that simple approximate linear models, [20], can be employed to design LTI controllers that give highly accurate, stable and robust HR control performance, e.g. [5, 21] (20 to 30 participants, mean RMSE below 3 bpm).
To directly address the HRV disturbance, a HR control approach was developed that allows the frequencydomain characteristics of the closedloop inputsensitivity function, which is the transfer function from the HRV disturbance to the control signal, to be appropriately shaped [5]; for treadmill exercise, HR control was accurate (mean RMSE of 3.0 bpm, n = 30) and the control signal was smooth and stable (average power of changes in the control signal was low). Using this design approach as a foundation, and based on the observation that HR dynamics are not significantly different between treadmills and cycle ergometers, [22], a common control strategy was derived and experimentally tested with these two exercise modalities; it was found to give accurate tracking (mean RMSE of 3.1 bpm vs. 2.8 bpm, cycle ergometer vs. treadmill; n = 25) and low control signal intensity [11].
The primary contribution of the present work is, for the first time, the application of the inputsensitivityshaping approach for feedback control of HR to cycle ergometer exercise and the systematic analysis of its performance and robustness in a large experimental test series. A secondary contribution is a comparison with alternative linear and nonlinear controllers based upon data available in the literature. A single LTI feedback compensator was calculated using a linear firstorder plant model. The aim of the work was to assess controller performance in several experimental scenarios using quantitative measures of tracking accuracy and control signal intensity (a total of 73 feedback control experiments involving 49 individual participants were performed), and to analyse performance and stability robustness properties of the compensator using a large family of empirically derived plant models (73 individual plant models were used for the robustness analysis).
2 Methods
2.1 Control design
The lumped linear plant model, Eq. (1), is theoretically valid for smallsignal deviations around a nominal operating point; under this approximating condition, the HRV disturbance term d acts as an additive output disturbance independent of the control signal u. Furthermore, it is recognised that different levels of exercise intensity, as characterised by HR, will lead to different levels of HRV [19].
The reason for the strictly proper constraint on C_{fb} is that the resulting inputsensitivity function U_{o}, Eq. (6) below, is also strictly proper as follows: as a consequence, both C_{fb} and U_{o} are low pass systems (i.e. HCode \(\lim _{\omega \rightarrow \infty } C_{\text {fb}}(j\omega ) = 0\) and \(\lim _{\omega \rightarrow \infty } U_{o}(j\omega ) = 0\)), thus making the feedback loop, and, in particular, the control signal u, insensitive to highfrequency noise and disturbances.
The reference prefilter C_{pf} was calculated to make the overall closedloop transfer function from reference r to output y equal to a standard secondorder system with critical damping and specified rise time ([5, 11]; here, the rise time was set to either 120 s [control tests with cohort A, see Section 2.4] or 150 s [cohort B]).
From Eq. 6, it is seen that the effect of the HR disturbance term d on the controller output signal u is governed by the inputsensitivity function U_{o}, which effectively acts as a filter for d. Thus, explicit shaping of the frequency response of U_{o} obviates the need for any separate filtering of the HR signal.
2.2 Materials
HR was recorded using a chest belt (T34, Polar Electro Oy, Finland) that communicated wirelessly with a receiver module (Heart Rate Monitor Interface [HRMI], Sparkfun Electronics, USA) connected via USB to the PC. The HR signal was interfaced to the Simulink model using a sample rate of 1 Hz. The feedback controller ran at a rate of 0.2 Hz (sample interval of 5 s); at each controller sample instant, the current HR value was taken as the mean of the latest five discrete HR samples.
The choice of controller sample interval of 5 s was based on formal guidelines for closedloop control systems: it is recommended that the controller sampling rate should be 10 to 30 times the closedloop bandwidth [23, page 110]. Since the chosen bandwidth for the closedloop inputsensitivity function is 0.01 Hz, the appropriate range for sampling frequency is 0.1 Hz to 0.3 Hz. Here, the sampling rate was chosen to be exactly in the middle of this range, viz. 0.2 Hz, corresponding to a sample interval of 5 s. Since the raw HR signal was available at a rate of 1 Hz, it was resampled to the appropriate controller sampling rate of 0.2 Hz as described above. The dynamic effect of this resampling is considered to be negligible, given that the plant time constant of 65.6 s is an order of magnitude higher than the controller sample interval of 5 s. Furthermore, the sampling rate of 0.2 Hz is more than a decade above the chosen closedloop inputsensitivity bandwidth of 0.01 Hz; this choice of sampling rate will therefore lead to no appreciable effects of the HR resampling on the closedloop system.
2.3 Outcome measures
For tests involving a squarewave HR target profile, RMSE and P_{∇u} were calculated over an evaluation period from 300 to 1800 s; when the target HR was constant, the evaluation period was from 400 to 1200 s.
2.4 Experimental procedures

Participant cohort A comprised 25 males (n = 25) with age on the range 22 years to 32 years, body mass from 62 to 114 kg, height from 1.65 to 1.93 m and body mass index (BMI) from 19.9 to 34.0 kg/m^{2}. This cohort previously participated in separate studies of system identification [22] and feedback control [11] using both a cycle ergometer and a treadmill. As noted above (Section 2.1), the nominal plant parameters k and τ in Eq. 1 that were used for controller calculation are average values obtained in the identification experiments with cohort A [22].

Participant cohort B had 24 males (n = 24) aged from 22 to 36 years, mass from 62 to 113 kg, height from 1.72 to 2.00 m and BMI from 18.8 to 32.5 kg/m^{2}. Two system identification series were conducted with cohort B: the first series (denoted B1) used the same experimental protocol applied to cohort A, [22], with a constant cycling cadence of 70 rpm; in the second series (B2), participants were allowed to cycle at their own preferred cadence, which was allowed to vary. During these tests, ratings of perceived exertion (RPE) were recorded four times at intervals of five minutes using the Borg RPE scale [2].

Cohort A (n = 25): a squarewave target HR profile was employed with a period of 10 min and variations of ± 10 bpm around an individual, moderatetovigorous intensity HR level denoted HRmid and calculated as described below. Participants maintained a constant cadence of 70 rpm by monitoring a visual display.

Cohort B (n = 24): the target HR was constant and equal to HRmid; cadence was constant at 70 rpm.

Cohort B (n = 24): target HR was again constant at HRmid, but participants were able to selfselect their preferred cadence.
For cohort B, the order of presentation of the two test conditions (constant or freely chosen cadence) was counterbalanced by changing the test order for consecutive participants.
The midlevel, moderatetovigorous intensity HRmid used to design the target HR profiles was calculated individually for each participant as follows. The boundary between exercise intensities perceived to be moderate and vigorous occurs at 76.5 % of maximal HR [1], and maximal HR can be estimated as HRmax (bpm) = 220 − age (years) [24]. Furthermore, it has been documented that, for similar levels of perceived exercise intensity, HR on a cycle ergometer is approximately 20 bpm lower than on a treadmill [25]. Thus, the midlevel HR target was set to HRmid (bpm) = 0.765 × (220 − age) − 20.
3 Results
3.1 Feedback control
The mean values of RMSE and P_{∇u} for the two cohort B conditions were found not to be significantly different: for RMSE, the p value was p = 0.45, and for P_{∇u}, p = 0.96 (cadence of 70 rpm vs. freely chosen cadence, pairedsamples twosided t tests, significance level α = 0.05).
3.2 Parametric plant model uncertainty
 1.
Participant cohort A (n = 25; full study details are given in [22]): using a sample of 25 healthy males, the steadystate gain and time constant were estimated for exercise on the cycle ergometer to be k = 0.392 bpm/W ± 0.120 bpm/W (mean ± standard deviation; range 0.180 bpm/W to 0.796 bpm/W) and τ = 68.7 s ± 21.5 s (range 38.1 s to 120.2 s), respectively. In this study, participants were required to maintain a constant pedalling cadence of 70 rpm.
(The nominal plant gain k = 0.392 bpm/W in Eq. 1 was taken as the mean value from this identification study. The nominal plant time constant τ = 65.6 s in Eq. 1 is the mean obtained across the 25 cycle ergometer values and 25 separate measurements with the same participants exercising on a treadmill, where τ = 62.5 s ± 18.5 s [range 34.3 to 110.1], see [22]).
 2.
Participant cohort B (n = 24): models for a separate sample of 24 healthy males were identified on the cycle ergometer using the same experimental protocol detailed in [22], with cadence = 70 rpm; this identification series is denoted B1. This resulted in the estimates k = 0.372 bpm/W ± 0.094 bpm/W (range 0.227 bpm/W to 0.565 bpm/W) and τ = 71.8 s ± 21.7 s (range 43.0 s to 133.2 s).
 3.
Participant cohort B (n = 24): this identification series, denoted B2, used the same cohort as in B1, but participants were allowed to freely choose their pedalling cadence. The outcomes were k = 0.364 bpm/W ± 0.074 bpm/W (range 0.197 bpm/W to 0.518 bpm/W) and τ = 65.7 s ± 24.2 s (range 26.5 s to 125.6 s).
The mean values of k and τ obtained from identification series A, B1 and B2 are deemed consistent because statistical testing revealed no significant differences between different pairs of conditions: τ, B1 vs. B2 (p = 0.30); k, B1 vs. B2 (p = 0.65); τ, A vs. B1 (p = 0.62); k, A vs. B1 (p = 0.53); τ, A vs. B2 (p = 0.65); k, A vs. B2 (p = 0.33). For the B1 vs. B2 comparisons, pairedsamples twosided t tests were conducted; for A vs. B1/B2, independentsamples twosided t tests were used.
For the series B1 and B2, there was no significant difference in mean RPE: 13.3 ± 1.4 vs. 13.1 ± 1.7, B1 vs. B2, p = 0.089 (pairedsample twosided t test).
There was a positive linear correlation between k and τ (r = 0.43, p = 0.00017; Fig. 6).
3.3 Robustness analysis
Performance and stability robustness properties of the feedback compensator C_{fb} in Eq. 4 were analysed using the family of 73 plant models obtained empirically as described above.
4 Discussion
The single linear compensator was found to give highly accurate HR tracking performance in both experimental cohorts and under the different experimental conditions: mean RMSE was on the range 2.5 bpm to 3.1 bpm. Due to the dynamic nature of squarewave reference tracking, mean RMSE for this condition (3.1 bpm) was higher than for the two constant target regulation series (2.5 bpm and 2.6 bpm).
The inputsensitivityshaping control design approach gives a simple, closedform analytical procedure that allows the closedloop bandwidth to be set in consideration of the broadspectrum HRV disturbance. In the present set of experiments, this gave a stable and smooth control signal whose changes had low average power P_{∇u} (mean of 2.4 W^{2} for constant HR regulation and 10.3 W^{2} for squarewave tracking).
Although the HR response was represented using the approximation of a simple linear model of the form y = P_{o}(s)u + d (Eq. 1, Fig. 1), where the term d represents the lumped effects of the HRV disturbance at a nominal operating point, it should be emphasised that human heart rate variability arises from complex interactions between the sympathetic and parasympathetic divisions of the autonomic nervous system [18]. These divisions are continuously engaged in regulation of cardiac output by adjustment of stroke volume and heart rate, thus leading to the observed variations in the time between individual beats. This HRV depends on many factors that are not dependent upon the control signal u (target work rate) including hydration level, ambient temperature and health status. Thus, it is not the purpose of the feedback control loop and, in particular, the control signal u, to directly influence the level of HRV. Rather, HRV is treated as a lumped, unmeasurable output disturbance d; the task of the controller is then, in the face of the unknown HRV disturbance d, to achieve a sufficient level of accuracy in the tracking of the target HR profile while maintaining an acceptable intensity of the control signal u. This amounts to the classical tradeoff between tracking accuracy and control signal intensity: choice of a higher closedloop bandwidth will tend to give a more dynamic controller resulting in lower tracking error but higher control signal intensity, and vice versa.
In comparison with the study of Kawada et al. [6], which employed a PI controller and evaluated only constant HR regulation, the RMSE values for regulation in the present work are slightly lower (\(\sim \)2.5 bpm here vs. 2.5 bpm to 3.8 bpm in [6]); but this comparison should be interpreted with caution since RMSE will also have been affected by the differing experimental conditions and the respective methods for controllerparameter tuning.
A direct comparison of the intensity of control signal activity between the two studies is not possible: here, this was evaluated using the average power of sampletosample changes in the control signal P_{∇u}; but in [6], no quantitative assessment of control signal intensity was performed. It can be conjectured, however, that the control signal intensity when using a PI controller (as in [6]) would be higher. This is because, in the present work, the compensator, Eq. 2, was constrained at the outset to be strictly proper (low pass). This in turn gives a strictly proper, lowpass inputsensitivity function U_{o} because, from Eq. 6, U_{o} = C_{fb}/(1 + C_{fb}P_{o}). Thus, when \(\lim _{\omega \rightarrow \infty } C_{\text {fb}} = 0\), it follows that \(\lim _{\omega \rightarrow \infty } U_{o} = \lim _{\omega \rightarrow \infty } C_{\text {fb}} = 0\). Thus, the control signal will not respond to disturbances at frequencies above the specified inputsensitivity bandwidth p (set here as frequency f = 0.01 Hz; see U_{o} in Fig. 2).
In contrast, for a PI controller C_{fb}(s) = k_{p} + k_{i}/s with proportional gain k_{p} and integrator gain k_{i} (this is the exact structure employed in [6]), the magnitude of C_{fb} tends to the value k_{p} at high frequency. Consequently, U_{o} also tends to the value k_{p} because, employing the condition that P_{o} is strictly proper (low pass), \(\lim _{\omega \rightarrow \infty } U_{o} = \lim _{\omega \rightarrow \infty } C_{\text {fb}} = k_{p}\). This shows that, for a PI controller, the control signal will react to disturbance and noise inputs across the whole frequency spectrum.
Finally, in comparison with the study in [6], it is noted that the nominal plant gain used here for controller calculation (k = 0.39, mean from 25 participants) was very close to the value estimated in [6] (k = 0.42, mean from 10 participants). The nominal time constant used here cannot be compared because a nonparametric model was estimated in [6].
The performance of the controller proposed and tested in the present work can be compared with nonlinear strategies that have previously been applied to HR control. One nonlinear approach has been applied to HR control during both treadmill [4] and cycle ergometer [7] exercise. This nonlinear method is based upon a plant model where the control signal u appears in quadratic form, and where the controller cancels this term using the inverse nonlinearity, viz. the squareroot function. This approach has the important theoretical property that global convergence of regulation errors is guaranteed for the class of nonlinear models considered. However, the experimental evidence provided in [4] and [7] is weak because no quantitative measures of controller performance were employed, and because shortduration tests were performed with only two (treadmill, [4]) or three (cycle, [7]) participants. Furthermore, a later independent study systematically compared this nonlinear approach to a linear PI controller using quantitative outcome measures and a cohort of 16 healthy male participants during treadmill exercise [26]. Using formal statistical analysis methods, this study found no significant difference between the linear and nonlinear controllers in HR tracking accuracy (for both controllers, RMSE was approximately 2.3 bpm) and in average control signal power. Moreover, the nonlinear controller was found to be highly sensitive at low control signal levels, which was attributed to the fact that the squareroot function, which is included in the compensator, has a gain that tends to infinity as the control signal tends to zero.
The HR tracking accuracy reported in [26] for both the linear and nonlinear controllers, i.e. RMSE of approximately 2.3 bpm, is slightly lower than the range of 2.5 bpm to 3.1 bpm observed in the present work. This can likely be attributed to the nonstrictlyproper nature of the linear/nonlinear controllers implemented in [26], in contrast to the strictly proper constraint applied here (Eq. 2): when the controller is not strictly proper, its gain does not roll off with frequency, thus making it more dynamic across the whole frequency range, which tends to drive down the RMSE; the price to be paid for this improved HR tracking accuracy, however, is an increased sensitivity to higher frequency HRV disturbances and consequent higher average control signal power.
Notwithstanding this critical analysis of nonlinear control strategies, further work is recommended to investigate appropriate nonlinear plant model and controller structures, while experimental evaluations are recommended that comprise quantitative performanceoutcome measures and participant cohorts with sufficient sample size to allow formal statistical comparison with other linear/nonlinear approaches.
Within the present work, the quantification of parametric plant uncertainty showed that steadystate gains and time constants vary over a very wide range; overall, k was on the range 0.180 bpm/W to 0.796 bpm/W and τ ranged from 26.5 to 133.2 s (Section 3.2, Fig. 6). Despite this high level of plant dispersion, the controller was accurate and stable in all 73 experiments involving a total of 49 individual participants.

Performance robustness: the magnitudes of the inputsensitivity and sensitivity functions, U and S, respectively, were found to be almost entirely unaffected by the plant variability at frequencies above the selected closedloop bandwidth p (which corresponds to f = 0.01 Hz, Fig. 7a), i.e. in the frequency range that is important in relation to the behaviour of the control signal. Furthermore, the complementary sensitivity function magnitude T was found to be little affected at the lower end of the ultra low frequency band (Fig. 7b), i.e. at frequencies that are primarily important for reference tracking accuracy.

Stability robustness: very large stability margins were evident across the whole family of plant models (Fig. 8): gain margin was infinite in all cases while the minimum phase margin remained large at 62.2^{∘} (nominal phase margin was 81.2^{∘}).
5 Conclusion
The singlelinear, timeinvariant controller was found to give accurate and stable performance with low values of the quantitative outcomes rootmeansquare tracking error RMSE and average control signal power P_{∇u}. The empirical evidence of controller robustness was corroborated by numerical analysis of key closedloop transfer functions and stability margins across the available plant model family.
These results, taken together with data reported in Kawada et al. [6], demonstrate that highly accurate, stable and robust heart rate control can be achieved using LTI controllers of remarkably simple structure. Furthermore, the results highlight that design methods for HR control must give adequate attention to plant disturbances caused by physiological heart rate variability. The inputsensitivity approach evaluated in this work provides a direct and transparent method of addressing this challenge.
Notes
Acknowledgments
Reto Grunder and Andreas Zahnd contributed to data acquisition and analysis in relation to the preceding system identification [22] and feedback control [11] studies, as noted in the text.
Author contributions
KH designed the study. CH did the data acquisition. CH and KH contributed to the analysis and interpretation of the data. KH wrote the manuscript and CH revised it critically for important intellectual content. Both authors read and approved the final manuscript.
Funding disclosure
This research did not receive any specific grant from funding agencies in the public, commercial, or notforprofit sectors.
Compliance with ethical standards
Conflict of interest
The authors declare that they have no conflict of interest.
Ethical approval
All procedures performed in this study involving human participants were in accordance with the ethical standards of the Ethics Committee of the Swiss Canton of Bern (Ref. 201701894) and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Full ethical approval was obtained.
Informed consent
Informed consent was obtained from all individual participants included in the study.
Consent for publication
Written informed consent for publication was obtained from the individual depicted in the image in Fig. 3.
References
 1.Riebe D, Ehrman JK, Liguori G, Magal M (eds) (2018) ACSM’s guidelines for exercise testing and prescription, 1st edn. Philadelphia, Wolters KluwerGoogle Scholar
 2.Borg G (1990) Psychophysical scaling with applications in physical work and the perception of exertion. Scand J Work Environ Health 16(Suppl. 1):55–58CrossRefGoogle Scholar
 3.Kawada T, Sunagawa G, Takaki H, Shishido T, Miyano H, Miyashita H, Sato T, Sugimachi M, Sunagawa K (1999) Development of a servocontroller of heart rate using a treadmill. Jpn Circ J 63:945–950CrossRefGoogle Scholar
 4.Scalzi S, Tomei P, Verrelli CM (2012) Nonlinear control techniques for the heart rate regulation in treadmill exercises. IEEE Trans Biomed Eng 59(3):599–603CrossRefGoogle Scholar
 5.Hunt KJ, Fankhauser SE (2016) Heart rate control during treadmill exercise using inputsensitivity shaping for disturbance rejection of verylowfrequency heart rate variability. Biomed Signal Process Control 30:31–42CrossRefGoogle Scholar
 6.Kawada T, Ikeda Y, Takaki H, Sugimachi M, Kawaguchi O, Shishido T, Sato T, Matsuura W, Miyano H, Sunagawa K (1999) Development of a servocontroller of heart rate using a cycle ergometer. Heart Vessels 14(4):177–184CrossRefGoogle Scholar
 7.Paradiso M, Pietrosanti S, Scalzi S, Tomei P, Verrelli CM (2013) Experimental heart rate regulation in cycleergometer exercises. IEEE Trans Biomed Eng 60(1):135–139CrossRefGoogle Scholar
 8.Argha A, Su SW, Celler BG (2017) Heart rate regulation during cycleergometer exercise via eventdriven biofeedback. Med Biol Eng Comput 55(3):483–492CrossRefGoogle Scholar
 9.Aseltine RG, Feldman CL, Paraskos JA, Moruzzi RL (1979) A simple device for closed loop heart rate control during cardiac rehabilitation. IEEE Trans Biomed Eng BME26(8):456– 464CrossRefGoogle Scholar
 10.Tennant LK, Allen RE, Pollock ML, Graves JE, Carmichael M, Conti R (1989) Evaluation of a heart rate controlled cycle ergometer for use in cardiac rehabilitation. J Cardiopulm Rehabil 9(5):195–201CrossRefGoogle Scholar
 11.Hunt KJ, Zahnd A, Grunder R (2019) A unified heart rate control approach for cycle ergometer and treadmill exercise. Biomed Signal Process Control 54CrossRefGoogle Scholar
 12.Su SW, Chen W, Liu D, Fang Y, Kuang W, Yu X, Guo T, Celler BG, Nguyen HT (2010) Dynamic modelling of heart rate response under different exercise intensity. Open Med Inform J 4:81– 85CrossRefGoogle Scholar
 13.Zhang Y, Haddad A, Su SW, Celler BG, Coutts AJ, Duffield R, Donges CE, Nguyen HT (2014) An equivalent circuit model for onset and offset exercise response. BioMedical Engineering OnLine 13:145CrossRefGoogle Scholar
 14.Cheng TM, Savkin AV, Celler BG, Su SW, Wang L (2008) Nonlinear modeling and control of human heart rate response during exercise with various work load intensities. IEEE Trans Biomed Eng 55(11):2499–2508CrossRefGoogle Scholar
 15.Su SW, Huang S, Wang L, Celler BG, Savkin AV, Guo Y, Cheng TM (2010) Optimizing heart rate regulation for safe exercise. Ann Biomed Eng 38(3):758–768CrossRefGoogle Scholar
 16.Su SW, Wang L, Celler BG, Savkin AV, Guo Y (2007) Identification and control for heart rate regulation during treadmill exercise. IEEE Trans Biomed Eng 54(7):1238–1246CrossRefGoogle Scholar
 17.Nguyen TN, Su S, Celler B, Nguyen H (2014) Advanced portable remote monitoring system for the regulation of treadmill running exercises. Artif Intell Med 61(2):119–126CrossRefGoogle Scholar
 18.Malik M (1996) Heart rate variability. Standards of measurement, physiological interpretation, and clinical use. Eur Heart J 17(3):354–381CrossRefGoogle Scholar
 19.Hunt KJ, Saengsuwan J (2018) Changes in heart rate variability with respect to exercise intensity and time during treadmill running. BioMedical Engineering OnLine 17:128CrossRefGoogle Scholar
 20.Hunt KJ, Fankhauser SE, Saengsuwan J (2015) Identification of heart rate dynamics during moderatetovigorous treadmill exercise. BioMedical Engineering OnLine 14:117CrossRefGoogle Scholar
 21.Hunt KJ, Gerber S (2017) A generalised stochastic optimal control formulation for heart rate regulation during treadmill exercise. Systems Science & Control Engineering 5(1):481–494CrossRefGoogle Scholar
 22.Hunt KJ, Grunder R, Zahnd A (2019) Identification and comparison of heartrate dynamics during cycle ergometer and treadmill exercise. PLoS ONE 14(8):e0220826CrossRefGoogle Scholar
 23.Åstrȯm KJ, Wittenmark B (2011) Computer controlled systems: Theory and design, 3rd edn. Dover Publications, MineolaGoogle Scholar
 24.Shargal E, KislevCohen R, Zigel L, Epstein S, PilzBurstein R, Tenenbaum G (2015) Agerelated maximal heart rate: examination and refinement of prediction equations. J Sports Med Phys Fitness 55:1207–1218PubMedGoogle Scholar
 25.Roecker K, Striegel H, Dickhuth HH (2003) Heartrate recommendations: transfer between running and cycling exercise?. Int J Sports Med 24:173–178CrossRefGoogle Scholar
 26.Hunt KJ, Maurer RR (2016) Comparison of linear and nonlinear feedback control of heart rate for treadmill running. Systems Science & Control Engineering 4(1):87–98CrossRefGoogle Scholar
Copyright information
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.