Mechanics of six-legged runners

Insect gaits and stepping patterns have been described extensively, because insects are used as models for the study of neural function (Delcomyn, 1981). However, few studies have directly addressed the mechanics of six-legged locomotion. The kinetics and kinematics of locomotion have been used primarily as a tool for understanding the interaction of peripheral force inputs and the neural control of locomotion (Delcomyn, 1985; Cruse, 1976). Graham (1983) suggested that the walking behavior of stick insects, as well as of other insects, is fundamentally different from that of birds and mammals. Stick insects simultaneously exert braking and propulsive forces during part of each stride. At certain periods during the leg cycle, the front or middle legs actively resist the forward motion of the body, producing a ‘lurching’ type of locomotion which may preclude mechanisms of energy exchange and transfer used in animals with different numbers of legs.

Despite differences in body size, body form, type of skeleton, leg number and leg orientation, animals as diverse as crabs, birds and mammals exhibit remarkable similarities in their locomotor mechanics. Quadrupedal and bipedal mammals, bipedal birds and eight-legged, sideways-running ghost crabs produce similar patterns of force and mass-specific energy changes during locomotion (Cavagna et al. 1977; Fedak et al. 1982; Heglund et al. 1982a; Blickhan and Full, 1987). These remarkable similarities reflect the common use of two basic mechanisms for minimizing energy expenditure by alternately storing and recovering energy. The first mechanism operates during walking. As the animal’s center of mass rises and falls, kinetic and gravitational potential energy are exchanged rather than lost, in a manner analogous to a swinging pendulum or an egg rolling end over end (Cavagna et al. 1976, 1977; Heglund et al. 1982a). While hopping, trotting and galloping, animals use a second mechanism analogous to what occurs in a bouncing ball. In large animals, elastic elements such as tendons alternately store and release energy during each stride (Cavagna et al. 1964; Heglund et al. 1982b). Not only do birds, mammals and ghost crabs appear to use similar mechanisms of locomotion, but they also perform similar amounts of mechanical work per unit of mass (Heglund et al. 1982b; Blickhan and Full, 1987).

The results obtained from studies on ghost crabs indicate that the energyconserving mechanisms which operate in vertebrates could be general features of legged locomotor systems. If this proves to be the case, then proposed explanations of mechanical similarity should not be based on morphology and physiology unique to birds and mammals. With data on only a single species of arthropod, the ghost crab, generalities may be premature. The locomotor mechanics of the most successful terrestrial animals, the insects, have not been examined. To address the question of whether fundamental patterns exist in the terrestrial locomotion of legged vertebrates and arthropods, we quantified the energetics of the center of mass of a six-legged insect, the cockroach, Blaberus discoidalis. This species was easily induced to run, and was of sufficient size to produce forces during locomotion which could be measured and resolved with our miniature force platform.

Blaberus discoidalis (mean mass 2.6 g; N=13) were obtained from Carolina Biological Supply Company. Cockroaches were housed in individual plastic containers with a layer of cedar shavings. Animals were given water and Purina Puppy Chow, ad libitum. All cockroaches were kept on a local photoperiod at ambient temperatures (24±2°C).

A track was constructed with Plexiglas walls to contain the animals during filming and force measurements. A cardboard floor gave the animals sufficient traction to prevent slippage. The force platform was inserted into the floor of the track, flush with the running surface and 50 cm from the start of the track. The animals ran into a darkened cardboard box placed 10–20 cm from the force platform.

We measured vertical and horizontal ground reaction forces using a miniature force platform based on a design by Heglund (1981). A platform measuring 10.7cm×6cm×0.06cm and weighing 3.07g was constructed of model aircraft plywood and mounted on four brass beams (Fig. 1). Semiconductor strain gauges were bonded to spring blades cut from the brass supporting beams at each corner of the platform. Cross-talk between vertical and horizontal outputs was less than 2%. Compensating gauges at each corner reduced drift due to temperature change. Loads in the range 0.001–0.1 N produced a linear response with a maximum variation across the platform of less than 7% (Fig. 2A). Test signals were digitally filtered using a Butterworth filter with zero phase shift. After repeated testing, we selected a cut-off frequency of 185 Hz (unloaded natural frequency of the platform=400–650 Hz). This cut-off frequency did not result in amplitude modulation or loss of any signal components. Fig. 2B shows the effects of filtering and the phase response of the force platform for a known signal generated by a pendulum.

To measure lateral forces, we rotated the force platform 90°. Gauges which measured force in the horizontal direction were then in position to measure lateral forces. Vertical and lateral forces were measured simultaneously for each trial. traction to prevent slippage. The force platform was inserted into the floor of the track, flush with the running surface and 50 cm from the start of the track. The animals ran into a darkened cardboard box placed 10–20 cm from the force platform.

We measured vertical and horizontal ground reaction forces using a miniature force platform based on a design by Heglund (1981). A platform measuring 10.7cm×6cm×0.06cm and weighing 3.07g was constructed of model aircraft plywood and mounted on four brass beams (Fig. 1). Semiconductor strain gauges were bonded to spring blades cut from the brass supporting beams at each corner of the platform. Cross-talk between vertical and horizontal outputs was less than 2%. Compensating gauges at each corner reduced drift due to temperature change. Loads in the range 0.001–0.1 N produced a linear response with a maximum variation across the platform of less than 7% (Fig. 2A). Test signals were digitally filtered using a Butterworth filter with zero phase shift. After repeated testing, we selected a cut-off frequency of 185 Hz (unloaded natural frequency of the platform=400–650 Hz). This cut-off frequency did not result in amplitude modulation or loss of any signal components. Fig. 2B shows the effects of filtering and the phase response of the force platform for a known signal generated by a pendulum.

To measure lateral forces, we rotated the force platform 90°. Gauges which measured force in the horizontal direction were then in position to measure lateral forces. Vertical and lateral forces were measured simultaneously for each trial.

Signals from each force platform channel were amplified (Vishay, Measurements Group), and collected by an analog to digital converter (C-100, Cyborg) interfaced with a computer (IBM PC/AT). Force signals were sampled at a frequency of 1000 samples s⁻¹. Integrations of force records were performed using data acquisition and analysis software (Discovery, Cyborg). Energy fluctuations were calculated using macros linked in electronic spreadsheets (Lotus).

Each trial was taped using a video camera (Video Logic CDR460) capturing 60 frames s⁻¹. Top and lateral views were filmed simultaneously by placing a mirror at a 45° angle above the track. A digital timer placed in the camera’s field enabled us to synchronize video and force records. Video frames were grabbed and the coordinates of the leg positions at each frame were digitized into a computer (IBM PC/AT) with a video analysis system (Peak Performance Technologies Inc.).

An animal’s speed was calculated from the video recordings by dividing the force platform length by the time taken to cross the platform. This value was checked independently with the force records by dividing the length of the force platform plus the body length of the animal by the duration of the vertical force recording. The two methods produced values in agreement to within 10 %. In the case of some very high-speed trials (>0.60 ms⁻¹), for which the video record was less precise for speed determination, only the latter method was used.

At speeds above 0.14 m s⁻¹, the maxima and minima of the vertical force recording clearly defined each stride period (i.e. one complete cycle of protraction and retraction of a leg). Synchronization of the video recordings with the force recordings showed that the stride periods determined by the force record agreed with those recorded on the video tape. At the slowest speeds, the vertical force patterns were less clear, and stride periods were determined by video analysis alone. Stride frequencies were calculated from the duration of the stride or strides selected from the trials used for energy analysis.

The horizontal and lateral velocity of the center of mass were calculated by integrating the horizontal and lateral force recordings, respectively. The vertical velocity of the center of mass was calculated from integration of the vertical force recording, minus the body mass. An additional integration of the vertical velocity yielded the vertical displacement of the center of mass. The animal’s average speed was used as the integration constant for the horizontal velocity of the center of mass. The integration constants for the vertical and lateral component of velocity and for the vertical displacement of the center of mass were assumed to be zero. Drift in the horizontal and vertical force signals was negligible (Fig. 3). Trials in which the animal changed direction or bumped into the walls of the track were rejected. Time segments from each trial were selected for analysis if they contained an integral number of strides and if the sum of the increases and decreases in speed was within 10% of the animal’s average speed.

Calculations of the energy fluctuations of the center of mass were performed as in Blickhan and Full (1987), Cavagna (1975) and Heglund et al. (1982a). The horizontal, lateral and vertical kinetic energies of the center of mass (E_k) were calculated from the equation: ⁠, where v is the horizontal, lateral or vertical velocity and M_b is the body mass. The gravitational potential energy of the center of mass (E_p) was calculated from the vertical displacement (h) and the gravitational acceleration (g) according to the equation: E_p = M_bgh. The total energy of the center of mass (E_cm) was the sum of the vertical and horizontal kinetic energies and the gravitational potential energy at each sampling period (1ms). The power needed to lift and accelerate the center of mass (Ėcm) was calculated from the sum of the positive increments of E_cm over an integral number of strides, divided by the duration of the strides.

As a cockroach moved over the force platform, distinct maxima and minima were apparent in the ground reaction forces (Fig. 3). Each vertical force maximum was correlated with a step observed on the video tape. The period of two vertical force cycles was correlated with a stride period (i.e. the length of time required to complete one leg cycle). Video analysis showed that cockroaches used an alternating tripod gait at speeds from 0.08 to 0.66m s⁻¹. The right front, left middle and the right hind legs all moved simultaneously in a step, while each moved out of phase with its contralateral pair (Fig. 4A). Maximal horizontal accelerating force occurred near the end of retraction. Maximal horizontal decelerating force occurred as one set of legs contacted the ground following protraction, while the other set began retraction. Over most of the speed range tested, the summed forces of the legs produced two discrete maxima on the vertical recordings during each stride (Fig. 4B, t₁ and t₃). Vertical force decreased to below body weight (Fig. 4B, t₂) as the center of mass rose (Fig. 4F) following upward acceleration by the legs (Fig. 4D). As the center of mass fell, the vertical force increased. Horizontal force increased as the legs accelerated the center of mass forward, and decreased as the legs produced braking forces at the end of each step (Fig. 4C,E). Below 0.14m s⁻¹, the movements of the three legs comprising each alternating tripod were not as tightly coupled as during faster speeds, resulting in small vertical force maxima which corresponded to discrete steps by individual legs. The overall pattern of horizontal force, however, did not change as a function of speed.

Lateral force fluctuations were less than half the magnitude of the vertical force fluctuations (Fig. 5). Because lateral forces were less regular than vertical or horizontal forces, we performed a Fourier transform on vertical and lateral force signals that had been recorded simultaneously. Both vertical and lateral force patterns showed the greatest spectral magnitude at one frequency over the range of speeds measured. The ratio of vertical to lateral signal frequency was 1.8±0.1 (S.D.) for 20 trials. The period of the lateral force signal was twice that of the vertical signal. Each lateral force maximum or minimum was in phase with the maxima of vertical force (Fig. 5B, C). When an animal accelerated forwards and upwards, it also generated a lateral force which pushed it towards the side with only a single leg of the tripod on the ground (Fig. 5A, with arrows, and B). The maximum lateral force showed some variation in phase with respect to the vertical force. The phase relationship depended on whether the animal first pushed to the side, and then pushed upwards and forwards, or whether it pushed upwards and forwards and then to the side.

Fig. 6A shows the phase relationship between horizontal kinetic and gravitational potential energy (Fig. 4H, I). Both signals were considered to be periodic functions that repeat every step (i.e. every half of the stride period shown in Fig. 4 from time t=0 to t₂). If the maximum of gravitational potential energy for one step was in phase with the nearest maximum of horizontal kinetic energy, we defined this as a zero phase shift. If the nearest maximum of horizontal kinetic energy preceded the nearest maximum of gravitational potential energy, the shift was considered negative. If the nearest maximum of horizontal kinetic energy followed the maximum of gravitational potential energy, the shift was considered positive. If the horizontal kinetic and gravitational potential energy were completely out of phase, the phase shift would equal ±180°. The mean phase shift was −7.6±7.6 (S.E.) and was not significantly different from zero. Horizontal kinetic and gravitational potential energy were in phase. Phase shift was not a significant function of speed (F(_1,46)=4.0, P>0.05; r²=0.08, therefore the slope, β, was zero).

Stride frequency increased linearly with speed from 0.08 m s⁻¹ to approximately 0.35 ms⁻¹ (Fig. 7). Above this speed, stride frequency attained a maximum. Increases in speed above 0.35 ms^-1 were achieved by increasing stride length.

The mass-specific power used to lift and accelerate the center of mass (Ė_cm, W kg⁻¹) increased linearly with speed (v, ms⁻¹): Ė_cm=0.89±0.097v (S.E.) −0.029 (Fig. 8). At speeds between 0.08 and 0.23 m s⁻¹, gravitational potential energy and horizontal kinetic energy changes were approximately equal in magnitude. Fluctuations in lateral kinetic energy were more comparable to fluctuations in vertical kinetic energy, but less than one-tenth that of changes in potential energy (Fig. 5D, E, F). The contributions of vertical and lateral kinetic energy fluctuations to the total energy changes of the center of mass were negligible. As speed increased, the mechanical power used to accelerate the center of mass forward increased, whereas the mechanical power used to accelerate the center of mass upwards did not show a systematic change with speed. Consequently, at the fast speeds, the total mechanical power was determined primarily by the fluctuations in horizontal kinetic energy.

Over nearly the whole speed range tested, B. discoidalis utilized an alternating tripod stepping pattern similar to the symmetrical running gaits of other insects (Hughes, 1952; Delcomyn, 1971), eight-legged ghost crabs (Blickhan and Full, 1987) and trotting four-legged vertebrates (Fig. 4A). Two sets of proplusors moved the body. Cockroaches have been described as using a speed-independent ‘walking’ gait at intermediate and fast speeds and a speed-dependent ‘walking’ gait at very slow speeds (Delcomyn, 1971; Kozacik, 1981). However, Spirtio and Mushrush (1979) found speed independence at very low stride frequencies. In some cases, the transition between these ‘gaits’ may be graded (Hughes, 1952). Hughes (1952) concluded that cockroach ‘walking’ and ‘running’ are indistinguishable.

In the present study we used stepping pattern, stride frequency and mechanics to define walking and running. Data on B. discoidalis suggest that cockroaches used a running gait at nearly all speeds. We found no evidence of a walking gait in which a pendulum-like energy exchange was used, even at stride frequencies as low as 3 Hz. At very slow speeds, locomotion became intermittent, and only rarely did cockroaches sustain a constant average speed for more than a few strides. Similar results have been reported for mammals such as squirrels and chipmunks (Heglund et al. 1982a).

Pendulum-like energy exchange can characterize walking gaits. Analysis of whole-body mechanics in walking birds and mammals has revealed that horizontal kinetic energy and gravitational potential energy fluctuate out of phase, allowing energy transfer (Cavagna et al. 1977). Energy transfer or recovery by walking birds and mammals reduces the energy required to sustain a constant speed by 55–70% Cavagna et al. 1977; Heglund et al. 1982a). Pendulum-like energy exchange is not restricted to two-and four-legged walkers. We have discovered similar results in a sideways travelling, eight-legged ghost crab (Blickhan and Full, 1987). Horizontal kinetic energy and gravitational potential energy in B. discoidalis fluctuated in phase (Fig. 6A). The low recovery values indicated that cockroaches did not utilize an effective pendulum-like mechanism of energy conservation at any speed (Fig. 6B).

Cockroaches used a running gait over the entire speed range we observed. These six-legged runners produced patterns of ground reaction forces that are virtually indistinguishable from those produced by trotting eight-legged crabs and four-legged mammals and running bipeds (Fig. 4B,C; Cavagna et al. 1964, 1977). The symmetrical oscillation of vertical force (i.e. duration and amplitude of vertical force above and below body weight) supports our hypothesis that cockroaches were running or bouncing with a step frequency near their natural frequency (Cavagna et al. 1988). The shape of the vertical force (F_v) pattern can also be described at any time Tby a modified Fourier series (Alexander and Jayes, 1978), where F_v=a[cos(πt/T)−qcos(3π t/T)] from the time a foot is set down (−T/2) to when it is lifted (+T/2). For B. discoidalis, the shape factor describing the force pattern (q=−0.1 to 0) and duty factor (the fraction of a stride duration a foot is on the ground, β=0.5) corresponded to a trot when compared with two-and four-legged vertebrates.

The time course of mechanical energy fluctuations is also a defining feature of vertebrate running gaits. During running or bouncing gaits, horizontal kinetic and gravitational potential energy of the center of mass fluctuate in phase. Based on these relationships of energy transfer and leg movement, we found that eightlegged ghost crabs show a transition from walking to slow running or trotting (Blickhan and Full, 1987). By these criteria, the alternating tripod gait used by B. discoidalis was a run, similar to the bouncing gaits of two-, four-and eightlegged animals (Fig. 6; Blickhan and Full, 1987; Cavagna et al. 1977; Heglund et al. 1982a). Blaberus discoidalis did not attain an aerial phase although, at the fastest speeds, vertical force decreased to less than one-fifth of the animal’s weight. The absence of an aerial phase does not exclude the gait from being a run or a trot. Human subjects running with increased knee flexion adopt a gait in which vertical displacement of the center of mass decreases and the aerial phase is reduced to zero (McMahon, 1985).

Results from our present study on cockroaches suggest that insect locomotion is fundamentally similar to locomotion in two-and four-legged vertebrates. However, one investigator’s study of ‘walking’ in stick insects reached a different conclusion. Stick insects walk using ‘lurching’ locomotion during which legs simultaneously exert braking and propulsive forces during part of the stride cycle (Graham, 1983). The resulting velocity changes of the body relative to the leg movements are unlike those which characterize vertebrate walking. Patterns of horizontal velocity fluctuations seen in cockroach locomotion are similar to those described in stick insects, yet cockroaches attained high speeds. The pattern of horizontal velocity fluctuations is actually similar in cockroaches and running, as opposed to walking, vertebrates. The use of lurching gaits by insects would be expected to increase metabolic costs greatly (Graham, 1983). However, locomotor economy in B. discoidalis is not significantly different from that of other pedestrians of a similar mass (Herreid and Full, 1984; R. J. Full and A. Tullis, in preparation). Graham (1983) hypothesized that the type of walking behavior observed in stick insects may be necessary for insects to maintain static stability. Our preliminary data indicate that, at high speeds, cockroaches may not rely on static stability, but use dynamic adjustments instead. Rather than being a general feature of insect locomotion, lurching gaits may be a phenomenon restricted to slow walking speeds, or peculiar to stick insects.

As quadrupedal mammals increase speed, they typically switch from a trot to a gallop. Further increases in speed are achieved by increasing stride length, while stride frequency remains relatively constant. Among mammals, both speed and maximum sustainable stride frequency at the trot–gallop transition scale predictably as a function of body mass (Heglund et al. 1974). Smaller mammals switch from a trot to a gallop at lower speeds and higher stride frequencies than larger mammals. We have shown that eight-legged ghost crabs follow this same trend (Blickhan and Full, 1987). Surprisingly, B. discoidalis attained a maximum sustainable stride frequency as it increased speed (Fig. 7). To our knowledge this is the first demonstration of a maximum sustainable stride frequency in insects, although data from Delcomyn (1971) suggest the possibility for another species of cockroach. The maximum sustainable stride frequency and the speed at which it was attained scaled as predicted from data on quadrupedal mammals (Heglund et al. 1974).

Patterns of stride frequency, speed and ground reaction force for six-legged B. discoidalis, eight-legged crabs and two-and four-legged vertebrates suggest that animals representing a great diversity of body forms may operate in a similar manner, while running at ‘comparable’ speeds (Full, 1989). For the first time, quantification of stride frequency, external force production and energy fluctuations may allow comparisons of ‘equivalent’ gaits in very different morphological designs. The following simple calculation using a scaling model supports this possibility.

The hypothesis of dynamic similarity for terrestrial locomotion predicts that animals of different size will move in a similar fashion whenever they move at speeds proportional to the square roots of their effective leg length or hip height (Alexander and Jayes, 1983). Dynamic similarity is achieved when Froude numbers (the ratio of inertial to gravitational forces, u²/gl, where u represents speed, g is acceleration due to gravity and l is a general characteristic of length such as hip height) are equal. Based on its Froude number (2–3 where l=0.4cm), B. discoidalis is predicted to show a gait change between 0.3 and 0.4 ms⁻¹. This corresponds to the speed at the maximum sustainable stride frequency and is the speed at which quadrupeds change from a trot to a gallop. Precise definition of a gait change in B. discoidalis at the maximum sustainable stride frequency must await further kinematic analysis.

It appears that the observed patterns of force production, mechanical energy fluctuation and stride frequency or period could be described by a model based on an oscillating spring and mass system (R. Blickhan, unpublished results; Cavagna et al. 1988; McMahon, 1985). The fact that six-and eight-legged runners produce force patterns similar to those in mammals suggests a comparable whole-body stiffness. Animals may very well operate as tuned mechanical spring systems possessing similar spring-like characteristics (R. Blickhan, unpublished results; McMahon, 1985; Taylor, 1985). Similarities in mechanics may be due to similar material properties of the muscles or connective tissues.

The mechanical energy of the center of mass (E_cm) represents one component of the total mechanical energy generated during locomotion (Heglund et al. 1982b). The internal energy (E_i) produced to accelerate the limbs relative to the body, elastic strain energy (E_e) and the transfer of energy between segments add to the total mechanical energy of the system. The largest fraction of the energy appears to be involved in lifting and accelerating the center of mass, E_cm. However, in large fast-moving mammals E_t can equal or exceed E_cm (Heglund et al. 19826). The magnitude of E_e is unknown.

The rate of mass-specific mechanical work used to move the center of mass of a cockroach increased with speed and was not significantly different from that used by a mammal, bird or ghost crab moving at the same speed (Fig. 8). The amount of mechanical energy generated to move 1 kg of animal 1 m is approximately 1 J, regardless of body form. The contribution of E_i remains to be quantified. However, the combined limb-mass relative to the body is small for B. discoidalis, and trunk rocking appears to be minimal. Consequently, the contribution of E_i is likely to be smaller than the upper limit of 32% estimated for ghost crabs (Blickhan and Full, 1987). The magnitude of E_e for B. discoidalis is not known, although the ability of small animals to store and recover elastic strain energy may be limited (Biewener et al. 1981). Large mammals using bouncing gaits can alternately store and recover energy in elastic elements, such as tendons. Recovery of this elastic strain energy would reduce the mechanical energy that must be generated by muscles to maintain a constant running speed.

In summary, similarities in force production, stride frequency and mechanical energy production during locomotion suggest that there may be common design constraints for terrestrial locomotion, which scale with body mass and are relatively independent of body form, leg number and skeletal type.

We thank Reinhard Blickhan for his aid in designing our miniature force platform. Supported by NSF grant PCM87–19066, Berkeley Campus Biomedical Research Support grant (NIH 86–39) and a Presidential Undergraduate Fellowship to MST.