INTERNATIONAL JOURNAL OF SPORT BIOMECHANICS, 1992, 8, 152.164 Design and Evaluation of a Multi-Degree-of-Freedom Foot/Pedal lntirface for Cycling Dennis Wootten and Maury L. Hull Described is the design of a foot/pedal interface intended as a research tool in the study of overuse knee injuries in cycling. The interface enables the systematic variation of factors that may affect loads transmitted by the knee joint. It permits two degrees of freedom of movement, inversion/eversion and abduction/adduction rotations, either separately or in combination. The movement permitted by each degree of freedom can be either free or resisted by spring assemblies. Sample data were collected to demonstrate the function of the footlpedal interface. With no spring resistance, the interface functioned as intended by allowing free movement of the foot. Significant interaction was seen between the two degrees of freedom, with more motion and a larger absolute mean occurring when both degrees of freedom were allowed simultaneously. This emphasizes the need for a multi-degree-of-freedom interface when undertaking a comprehensive study of the factors affecting loads transmitted by the knee. Knee pain is perhaps cycling's most common overuse injury, accounting for approximately 25% of all reported cycling injuries (Gaston, 1977; Hannaford, Moran, & Hlavac, 1986). These ailments typically result from a combination of overuse and biomechanics (Pruitt, 1988). Pathomechanical cycling injuries are overuse injuries that result from incompatibilities between the bicycle and the bicyclist. Differences between the fixed motion patterns of the bicycle pedal and the natural motion of the cyclist's foot are thought to contribute to pathomechanical cycling injuries (Francis, 1988). During the power phase of cycling, when the foot is moving down it may attempt to pronate (Francis, 1986), which is a combination of dorsiflexion, abduction, and eversion (see Figure 1). If the foot cannot undergo this movement, then compensatory motions are developed in other components. Such motion may possibly be absorbed by the shoe and the slight play that exists with most conventional shoelpedal systems (Zahradnik, 1990). However, pronation may be restricted by stiff shoes and rigid shoelpedal connections, in which case the joints must absorb the motions and the resulting loads. There has been an increase in The authors are with the Department of Mechanical Engineering at the University of California-Davis, Davis, CA 956 16. Design of a Foot/Pedaf Interface Forefoot Abduction Forefoot Inversion X Forefoot Adduction Dorsifkxion Y Eversion Plantarf lexion Figure 1 - Orientation of axes and motions with respect to the right foot. The origin is located at the ball of the foot with positive X anterior, positive Y medial, and positive Z upward. Forefoot inversion/eversion are rotations about the X-axis, plantarflexion/dorsifIexion are rotations about the Y-axis, and forefoot adduction/ abduction are rotations about the Z-axis. injuries related to foot restraint as shoe uppers have become stiffer and strapless pedals have locked the foot more firmly in place (Zahradnik, 1990). In order to allow the foot complete angular mobility, three orthogonal axes of rotation are necessary. The orientation of these axes relative to the foot is described in Figure 1. Although they do not directly correspond with the anatomical axes of the joints of either the foot or the ankle, motions about these axes will allow unconstrained rotation of the foot while it is attached to the pedal. An axis parallel to the y-axis, which allows the foot to dorsiflex and plantarflex, is provided by the pedal spindle. Motions about the other two axes must be provided by the pedal interface. The ultimate goal of the research to which the present article pertains is to investigate whether a pedal that allows for joint motions accompanying pronation and supination will reduce the risk of pathomechanical knee injuries during cycling. Knee forces and moments, calculated from pedal forces and moments using a five-bar linkage kinematic model of the bicycle and bicyclist (e.g., Ruby, Hull, & Hawkins, 1991), will be evaluated as potential sources of overuse injury. It is assumed that reducing these forces and moments will benefit the cyclists by reducing both the risk of injury and fatigue. Pedal forces and moments will be measured by a six-load component pedal dynamometer with a pedal interface that either can allow for pronation and supination or can be locked Wootten and Hull Y Ctasp Hinge Spring Assemblies Linear Motion Potentiometer Concave Cylinder B. Inversion/Eversion Mod Axial ~ o t a t i o nModule n Connecting Plate is not visible Figure 2 - Pedal interface modules. The cleat connector provides the shoelpedal connection. The cleat connector bolts to the convex cylinder of the inversion/eversion module. The inversion/eversion module bolts to the axial rotation module through the interface plate. The limits of rotation are +lo0 for both modules. in a set position. The design of a pedal interface that allows the foot to pronate and supinate is the focus of this paper. Design Description The pedal interface consists of three modules: the cleat connector, the inversion1 eversion module, and the axial rotation module (see Figure 2). The cleat connector, Design of a Foot/Pedal Interface 155 shown in Figure 2a, provides the shoe/pedal connection. The interface uses a Shimano cleat that fits snugly into the front of the cleat connector. A hinged clasp rotates into the groove in the back of the cleat, and a pin locks the clasp securely in place, rigidly restraining the shoe. The inversion/eversion module, shown in Figure Zb, allows -t 10" of rotation, which occurs through a pair of mating aluminum partial cylinders sliding on each other. These surfaces are coated with Teflon-S to reduce friction and wear. The cleat connector bolts to the rorating convex cylinder which transfers downward forces to a mating, nonrotating concave cylinder below it. Thrust and upward forces are transferred to tabs bolted to the front and rear of the concave cylinder that fit into concave recessions in the convex cylinder. The clearances between these tabs and the convex cylinder are adjustable to achieve smooth rotation. A linear motion potentiometer is bolted through spherical bearings to the convex cylinder at one end and to an arm extending laterally from the front tab at the other end. Thus the potentiometer changes in length when the convex cylinder rotates. Restoring moments can be provided by spring assemblies mounted at the back of the module. The axial rotation module, shown in Figure 2c, allows 10" of forefoot abduction and adduction. The inversion/eversion module bolts to the interface plate, which in turn bolts to the axial rotation plate. Free to rotate, the axial rotation plate bears on the nonrotating cylinder housing below. Thus downward axial loads are supported by the bearing surface. Support of radial loads is achieved by bolting the axial rotation plate to a cylinder that fits into a mating hole in the cylinder housing. Arms extending from the bottom of this cylinder to the front and rear support upward axial loads, with the front arm actuating a spring assembly and the rear arm connecting to a linear motion potentiometer through a spherical bearing. The other end of the potentiometer is also bolted through a spherical bearing to an arm extending laterally and integral with the cylinder housing. Accordingly, the length of the potentiometer changes when the rear arm rotates. These parts are also coated with Teflon-S to facilitate the relative motion. Interconnecting the three modules not only enables the degrees of freedom offered by the individual modules but also enables adjustment of the position of the axes of rotation relative to the foot while maintaining the foot in a reference position relative to the pedal spindle. As indicated in Figure 3, the cleat connector is bolted to the inversion/eversion module through two slots that allow for 1.25 cm of medialflateral translation of the foot relative to the modules below. To obtain independent adjustments of the medialflateral position of both axes of rotation relative to the foot, the inversion/eversion module bolts to the interface plate of the axial rotation module below through slots in flanges on either end of the inversion/eversion module. Thus, adjusting the cleat connector mediallyflaterally moves the foot relative to the rotation axes of both modules, while subsequent medialflateral adjustment of the inversion/eversion module moves the foot relative to the axial rotation axis only. To obtain forelaft adjustment of the foot relative to the axial rotation axis, the interface plate bolts to the axial rotation plate through slots running forelaft. Repositioning the foot to a reference position relative to the pedal spindle following any or all of the above adjustments requires the capability for both forelaft and medialPatera1 recentering. This capability is realized through a Wootten and Hull Adjustment o f t h e Axial Rotation Axis Cleat Connector Fore/Af t Adjustment Inversion/Eversion Module F o r e / A f t Recentering MediaVLateral Recentering (Slot Not Visible) Figure 3 - Assembled modules. When assembled, the modules are interconnected to enable adjustment of the axes of rotation relative to the foot while maintaining the foot in a reference position relative to the pedal spindle. 1. Lock Nut 2. Spring Assembly Bracket 3. Piston Nut 4. Cylinder 5. Piston 6. Spring Actuator 7. Spring Figure 4 - Spring assembly cross-section. The spring assemblies provide restoring loads for both inversion/eversion and axial rotations. Each consists of a piston, piston nut, spring, cylinder, and lock nut. The piston slides in the cylinder with external loads resisted by a spring between the piston and cylinder. connecting plate between the axial rotation module and the dynamometer below. Forelaft recentering is achieved by bolting the axial rotation module through slots in flanges on either side of the plate, and medialPatera1 recentering is achieved through slots in flanges at either end of this plate where it bolts to the dynamometer. Thus the slots in the side flanges permit the axial rotation module to translate forelaft while the slots in the end flanges allow the connecting plate (with axial rotation module attached) to translate medially/laterally. The spring assemblies provide restoring moments for both inversion1 eversion and axial rotations. Diagrammed in Figure 4, each spring assembly consists of a lock nut, piston nut, cylinder, piston, and spring. The spring assembly Design of a FootIPedal Interface A. - 20 - 10 - Linear Spring B. Linear Spring with Clearance - 0 5 C E 157 0- a 3 -10 - -20 -4 -8 0 4 8 Displacement (mm) Displacement (mm) C. Preloaded Spring Displacement (mm) D. Preloaded Spring with Clearance Displacement (mm) Figure 5 - Spring load vs. displacement curves. Spring assemblies can be set to provide any of the four curves shown. can be configured to provide the spring curves depicted in Figure 5. Setting both the piston nut and the cylinder such that the spring is at its free length and the pistons just contact the actuator gives the standard spring curve shown in Figure 5a. Threading the cylinder out of the spring assembly bracket with the piston nut set so that the spring is at its free length creates a gap between the piston and the spring actuator, giving the load-versus-displacement curve shown in Figure 5b. Tightening the piston nut allows the spring to be preloaded. Preloading the spring and setting the cylinder so that the piston just contacts the spring actuator gives the curve shown in Figure 5c. Alternatively, preloading the spring and setting the cylinder so that a gap is created between the piston and the spring actuator gives the curve shown in Figure 5d. Spring adjustments can be done by hand with the exception of preloading the spring, which requires an allen wrench to hold the piston while the piston nut is tightened with a wrench. Threading a Wootten and Hull 158 bolt into the spring assembly bracket can impede the motion of the spring actuator, thus either reducing the range of rotation or locking the degree of freedom at a set angle. Calibration of Linear Motion Potentiometers A calibration was performed to relate the output voltages from the linear motion potentiometers to the pivot angles. Both potentiometers were calibrated over the full range of motion of 10" permitted by each of the modules. In the case of inversion/eversion, known angles were established by placing precision ground steel angles on top of the convex cylinder and pivoting it until the top surface of the steel angle was level. For the axial rotation potentiometer, known angles were established similarly except that the axial rotation module was placed on its side. To quantify the relations between the input angles and output voltages, these data for each potentiometer were used in a linear regression to determine the slope and intercept of the equation. + where V is the output voltage, b is the intercept, m is the slope, and @ is the input angle. These data proved to be very linear with an R2 of 0.9999. After the linear motion potentiometers were calibrated, an accuracy check was conducted. The angles were set using the methods described for the calibrations. The apparent angles were then calculated by solving the regression equations. The maximum absolute difference of the calculated angles minus the set angles was less than 0.2". Sample Angle Data Sample pedal angle data were collected to demonstrate the functions of the pedal interface. The cleat and axes of rotation were arbitrarily set in a neutral position with the cleat center located at a reference position of 6.35 cm lateral of the crank and 3.0 cm directly above the spindle center. The axial rotation axis was adjusted to pass through the cleat center and the inversion/eversion axis was adjusted such that it intersected the axial rotation axis. The degrees of freedom were allowed to free-float, without springs resisting the rotations. Data were collected for three test conditions: both degrees of freedom allowed, the axial axis locked in the middle of its range so that only inversion/eversion was allowed, and the inversion/eversion axis locked in the middle of its range so that only axial rotation was allowed. Axes were locked using screws in place of spring assemblies. The subject in this study was an experienced bicyclist who rode a racing bicycle mounted on a Schwinn Velodyne stationary trainer at a cadence of 92 rpm and a power output of 200 Watts. The rider was given approximately 5 minutes to adapt to each test condition before crank angle, pedal angle, and foottpedal interface angles were collected. Crank and pedal angles were measured following the procedures described in Hull and Davis (1981). Ten crank revolutions were collected using a personal computer equipped with an analog-todigital conversion interface. Data were sampled at a rate of 300 Hzlchannel. Footlpedal interface angles averaged over the 10 crank cycles are plotted Design of a Foot/Pedal Interface 159 versus the crank angle in Figures 6 and 7. The crank angle is the angle of the right crank viewed from a lateral position with clockwise (forward pedaling) being positive from the 0" reference at top dead center. The vertical bars represent the standard deviations over the 10 cycles. When only axial rotation was allowed (Figure 6), the average forefoot angle ranged from 0.8" of adduction to 1.2" of abduction with relatively large standard deviations that ranged from 1.1 to 1.9". Because the bicyclist's forefoot adducted on the downstroke and abducted on the upstroke with nearly equal extreme values, the average axial angle was only slightly abducted. The average axial rotation angle cycled once with every crank revolution. In contrast to axial rotation in which the overall average axial rotation was near O", the average inversion/eversion angle (Figure 6) remained everted and ranged from 2.7" of eversion to 7.1" of eversion, with an overall average foot position of approximately 4.5". Similar to the results for axial rotation, standard deviations were relatively large, ranging from 1.8 to 2.8". Also, the average inversion/eversion angle cycled once per crank revolution. Comparing the data when both degrees of freedom were allowed (Figure 7) with data when only axial rotation was allowed and when only inversion1 eversion was allowed (Figure 6) shows the presence of interactions between the two degrees of freedom. Although the curves for the same degree of freedom have similar shapes when either one or two degrees of freedom were allowed, the foot was more angled and had more angular motion when both degrees of freedom were allowed. The overall average eversion angle increased from 4.5 to 7.5" and the overall average axial angle went from being nearly 0 to 1.4" of + adduction Axial Rotation - abduction + inversion Inversion/Eversion - eversion Crank Angle (degrees) Figure 6 - Degree of freedom angle vs. crank angle when degrees of freedom are allowed individually. Averages over 10 crank cycles are plotted together with error bars indicating +1 SD. The average peak forefoot angles in abduction/adductionwere approximately equal so that the overall average rotation was nearly 0". The average inversion/eversion angles were always everted. Wootten and Hull 160 + adduction Axial - Rotation abduction + inversion InversionIEversion - 0 60 120 180 240 300 eversion 360 Crank Angle (degrees) Figure 7 - Degree of freedom angle vs. crank angle when both degrees of freedom were allowed. Averages over 10 crank cycles are plotted together with error bars indicating f1 SD. When both degrees of freedom were allowed, the overall averages and ranges of average values changed from when degrees of freedom were allowed individually, indicating interactions between the two degrees of freedom. adduction. The range of average axial rotation angle increased from 2.0 to 2.8" when inversion/eversion was allowed, while the range of average inversion1 eversion angle increased from 4.4 to 4.7" when axial rotation was allowed. Of the two range increases, that of inversion/eversionrotation is comparable to the measurement accuracy of the potentiometers. Accordingly, no significance should be attached to this result. Note that although the standard deviations in Figure 7 exceed the 10" limit of rotation, the extreme rotation actually measured was only -9.7". Thus the limits of rotation were never achieved by this subject. Discussion Pedal Interface Design The pedal interface designed is intended to be used as a research tool in the study of overuse injuries to the knee in cycling. Because many factors may affect the motion of the foot, and hence the loads developed at the knee joint, the pedal interface was designed to allow these factors to be varied systematically so that their effects might be investigated. One factor was the degrees of freedom. The interface offers two degrees of freedom, inversion/eversion and axial rotation, which can be used either separately or together. Although these axes do not correspond with anatomical axes, the interface will allow the foot to rotate about the anatomical axes while attached to the pedal, provided that both degrees of freedom are permitted simultaneously. The degrees of freedom can be permitted separately so that their Design of a FooVPedal Interface 161 effects can be isolated and studied. Isolating the effects is desirable for determining whether a simple footlpedal interface, with only a single degree of freedom, results in reduced knee loads. However, it was necessary to include both degrees of freedom to allow interactions between the degrees of freedom to be investigated. Another factor was the placement of the inversion/eversion and axial rotation axes relative to the foot. If the placement of the pedal interface axes of rotation coincides with the point of application of forces, then pivoting due to moments caused by these forces will be minimized. Therefore it may be desirable for the medialPatera1 and vertical placement of the inversion/eversion axis to correspond with the intersection of the center of vertical pressure and the center of medialPatera1pressure. This would minimize rotation due to these pedal forces while allowing for rotation due to applied moments, such as the foot attempting to rotate about the subtalar joint. Accordingly, the interface allows the inversionleversion axis position to be varied mediallyPaterally with respect to the foot so that the effect of this factor on knee moments may be investigated. The elevation of this axis is not adjustable, however, due to difficulties in designing an interface that permits changing the elevation of this axis while maintaining a constant height of the shoe sole above the pedal spindle. The elevation for this axis was chosen at the bottom of the shoe because this was believed to be approximately the center of medialPatera1 pressure. Axial rotation caused by forelaft and medialPatera1forces will be minimized in a similar manner if the axial rotation axis is located at the intersection of these two forces. Therefore the interface allows variation of the medialPatera1 and forelaft placement of the axis relative to the foot. However, the location of the applied loads may vary with crank angle and subject, making it impossible to establish ideal axis positions. Because the placement of the foot relative to the pedal spindle might affect pedal loads and ultimately knee loads, the interface included means for establishing a reference foot position once the axes are located. The reference position used in the tests in which the sample data were collected established the center of the cleat at a distance of 6.35 cm laterally from the crank, which is approximately 1.3 cm directly above the pedal spindle center. The extra lateral distance allows more axial rotation without the heel hitting the crank. The extra distance also provides increased clearance to accommodate medialPatera1 adjustments should these be performed. Because the elevation of the inversion/eversion axis cannot be adjusted in the present design, the reference position of the height of the shoe sole is established by the pedal dynamometer to which the interface attaches. An existing dynamometer (unpublished) to which the interface was mounted for the tests reported here established the height of the shoe bottom above the spindle center to be 3.0 cm, which is 1.0 to 1.5 cm higher than that established by many pedals. How the height affects results is unknown. Thus it would be desirable to connect the interface to a dynamometer that either establishes a conventional height or allows for height adjustment. The amount of pivoting provided by each degree of freedom is k 10". These limits were imposed by practical considerations. In the case of abduction/ adduction rotations, exceeding 10" abduction causes the medial malleolus to hit 162 Wootten and Hull the crank arm in the upstroke. Inversion/eversion rotation is limited by the inversion/eversion module rotating to the point where it bottoms out on the axial rotation module below. Although the inversion1eversion limit can be controlled through design, increasing the limit requires a compromise in height. To control the height, the limit was set at k 10". To investigate how different methods of applying restoring moments might affect knee loads, spring assemblies were designed to provide the four loaddisplacement relations depicted in Figure 5. The ability of the spring assemblies to provide these relations was confirmed through testing. The rationale behind the relation in Figure 5a is that spring force that varies linearly with displacement may be sufficient to reduce excessive pivoting. Permitting the foot to pivot until it reaches a certain angle before applying a spring force (Figure 5b) may be preferred if only a small amount of resistance is needed. Alternatively, it may be beneficial to preload the spring (Figure 5c) if greater restoring force is needed. Finally, the foot can also pivot freely until it reaches a certain angle before it hits a preloaded spring (Figure 5d) for a smoother deceleration than just hitting a stop. This would be useful for subjects whose natural motion would exceed the limits of rotation. In order to limit inertial effects caused by nonconstant angular crank velocities, the weight of the interface was an important concern. Varying angular crank velocity may be caused by noncircular chainrings, for example. A weight of 0.65 kg for the combined modules was achieved by utilizing the high strength and low density of 7075 T6 aluminum. The aluminum surfaces of the interface were coated with Teflon-S for bearings that have low friction and wear with only a slight increase in weight. Of course the total weight of the interfaceldynamometer assembly will depend on the weight of the dynamometer. A present dynamometer to which the interface can mount also weighs 0.65 kg, giving a total weight of 1.30 kg. Although this is about five times the weight of a typical pedal, test subjects do not notice any effect when pedaling with noncircular chainrings. Sample Results To demonstrate the function of the pedal interface, an experiment was conducted using a single subject who pedaled for three test conditions. Results of the experiment showed that the interface performed as desired for all test conditions. Rotations occurred smoothly without any evidence of either excess friction or binding. For the subject tested, motions were within the allowed ranges of rotation of 10". Hence the data collected appear to be a reasonable representation of the motions the foot would naturally undergo for each of the three test conditions. Because the interface performed as desired in permitting natural movement of the foot, it is worth discussing some of the key observations from the sample data. One is the relatively large standard deviations. Although the subject was an experienced cyclist, these standard deviations indicate between-cycle inconsistencies in the subject's pedaling mechanics. It is possible that, through increased adaptation time in using the equipment, these standard deviations would diminish. The sample results were remarkably different than expected. It was hypothesized that the bicyclist would pronate during the power phase. As the crank was + Design of a FootIPedal Interface 163 pushed down, the foot was expected to plantarflex, abduct, and evert (Francis, 1986). However, the sample results showed that the bicyclist's forefoot adducted (i.e., moved from being ab- to ad-ducted) and everted on the average during the power phase, which is not congruent with the normal motions of the foot and ankle (Root, Orien, & Weed, 1977). This unexpected motion of the foot could be a result of joint motion of the hip and knee as well as the ankle and foot, all of which act in concert to allow for adduction and eversion simultaneously. To explain this unexpected motion, it would be useful to monitor the kinematics of all joints in the chain. More motion was observed when both degrees of freedom were allowed simultaneously than when only a single degree of freedom was allowed. This could be due to the increased mobility of the ankle and foot joints as pronation and supination are allowed. The axes of the ankle and foot are all triplanar axes and any attempt to produce motion in only one plane will sublux and jam the joint (Root et al., 1977). Not allowing all three axes would reduce triplanar motion and load the joints. This interaction implies that footlpedal interfaces must include both degrees of freedom together to obtain accurate results pertinent to such interfaces. Studying each degree of freedom separately and superimposing the results would ignore these interactions and give misleading results. It is likely that interactions will be found in the loads as well as the motions. Conclusion The goal of this project was to design a pedal interface that allows the systematic variation of a number of factors that potentially affect loads transmitted by the knee joint and hence overuse injuries to this joint. A pedal interface was designed in three modules-the cleat connector, the inversion/eversion module, and the axial rotation module-which allow the foot f 10" of inversion/eversion and axial rotation. These degrees of freedom can be implemented either separately or together. The pedal interface allows the foot to be located in a reference position and the placement of the axes to be adjusted relative to this position. There are four spring configurations available to resist rotation. Sample pedal interface angle data were collected that demonstrate the functionality of the interface. With no spring resistance, the interface allowed free movement of the foot without excessive friction and within the 10"ranges permitted. Demonstrating an apparent significant interaction between the two degrees of freedom, angular motion increased when both degrees of freedom were allowed simultaneously compared to the motion occurring when each degree of freedom was allowed separately. Consequently, with its multi-degree-offreedom capability in addition to the many other design features, and with its demonstrated functionality, the interface in conjunction with an appropriate dynamometer and motion analysis system should provide the necessary instrumentation for investigating the effects of the various factors on knee loads. + References Francis, P.R. (1986). Injury prevention for cyclists: A biomechanical approach. In E.R. Burke (Ed.), Science of cycling (pp. 145-184). Champaign, IL: Human Kinetics. Francis, P.R. (1988). Pathomechanics of the lower extremity in cycling. In E.R. Burke Wootten and Hull (Ed.), Medical and scientific aspects of cycling (pp. 3-16). Champaign, IL: Human Kinetics. Gaston, E.A. (1977, November). Biker's knees. Bicycling. Hannaford, D.R., Moran, G.T., & Hlavac, H.F. (1986). Video analysis and treatment of overuse knee injury in cycling: A limited clinical study. Clinics in Podiatric Medicine and Surgery, 3, 671-678. Hull, M.L., & Davis, R.R. (1981). Measurement of pedal loading in bicycling: I. Instrumentation. Journal of Biomechanics, 14, 843-855. Pruitt, A.L. (1988). The cyclist's knee: Anatomical and biomechanical considerations. In E.R. Burke (Ed.), Medical and scientific aspects of cycling (pp. 17-24). Champaign, IL: Human Kinetics. Root, M.L., Orien, W.P., & Weed, J.H. (1977). Clinical biomechanics-Volume [I, Normal and abnormal function of the foot. Los Angeles: Clinical Biomechanics Corp. Ruby, T., Hull, M.L., & Hawkins, D. (1992). Three dimensional knee loading during seated cycling. Journal of Biomechanics, 25, 41-53. Zahradnik, F, (1990, July). Pivotal issue; Should your feet be fixed or floating? Bicycling, pp. 134-137. Acknowledgments We are grateful to the Shimano Corporation of Osaka, Japan, and Shimano American Corporation for providing the financial support for this project. We also thank Don Barnum of Mechanical General Services in the College of Engineering at UC-Davis for his excellent fabrication of the footlpedal interface.