This project is funded by a grant from The Leverhulme Trust. The work is being conducted by Robin Crompton, Todd Pataky and Paolo Caravaggi at Liverpool in collaboration with Bill Sellers (University of Manchester) and Tony Pridmore and Alistair Ross (University of Nottingham).
Human ancestors, or hominins, have been bipedal for at least four and a half million years. The feet of Ardipithecus already show adaptation for a toe-off mechanism that can have little function in other than terrestrial bipedalism. At 6 to 7 million years apes like Orrorin and Sahelanthropus which must lie close to the last common ancestor of humans and chimpanzees were already moving in an upright posture, and even the earliest recognizable relatives of living apes, at 11-21 MYA, show adaptations to upright trunk posture. But not till 1.8 million years ago were our ancestors striding, long distance walkers. What happened in between?
Hominin material from between 2 and c. 3.8 MY includes Australopithecus afarensis, eg. AL-288-1 (‘Lucy’) and DIK-1-1 (‘Lucy’s baby’). This species was almost unquestionably bipedal but with short legs, wide pelvis and long arms. There are two very contrasting views of the locomotion of Australopithecus afarensis AL-288-1, ‘Lucy’, still the best known early hominid. The first argues that because of skeletal features which suggest partial arboreality, that her bipedalism would have been a ‘compromised’ ,‘shuffling, ‘bent-hip, bent-knee’ bipedality, like that characteristically displayed by other African apes. The alternative sugests thatLucy was an habitual, upright, terrestrial biped.
When humans walk normally, the forces they exert against the ground show a characteristic double-humped pattern, indicating that the kinetic and the potential energies of the body centre of mass are fluctuating out of phase with each other, allowing a pendulum-like saving of energy. This is associated with pressure propagating from under the heel, down the lateral side of the foot, and, as the foot everts and pronates, across the ball of the foot to the big toe for push-off. In chimpanzees, the flexed knees and hips characteristic of their bipedal walking lead to a flat force curve, with KE and PE in phase and little or no energy savings. This is associated with peak pressure in the midfoot and no push-off from the big toe. In which way did early human ancestors move?
We have very few foot bones in the fossil record, and they tell us little about how the foot functioned, being described as ‘human-like’ in some features, ‘ape-like’ in others. But what we do have is footprints, which are our most direct evidence of the way our early ancestors walked. The best known set is the 3.5-3.8 MYA trails at Laetoli site G, in Tanzania. The only early human skeletons we find at Laetoli represent Australopithecus afarensis. Two adults, probably both females, from the size of the footprints compared to known footbones for male A. afarensis, walked one in front of the other, with a child walking alongside, across damp, fresh volcanic ash, for about 20 full strides.
With this number of strides, we can use the spacing of the footprints to predict speed, as long as we know stature and limb segment length. Unfortunately, as can be seen from the figure above, at some periods several species of bipedal walkers existed at the same time. However, today there is only one habitual bipedal walker, ourselves. Therefore, we can’t rely on analogy to reconstruct their behaviour, and how it differed from our own.
However, since F = M x A, we can predict the forces required or engendered within any system of rigid bodies – such as a skeleton, from its motion. We can therefore use computer simulation to predict forces from hypothesized motions and mass distributions (inverse dynamics) or motions from hypothesized (muscle) forces and mass distributions (forwards dynamics).
Some time ago we carried out inverse-dynamics computer simulations to determine which of these two gaits, erect or ‘bent-hip, bent-knee’ (BHBK) would have been most compatible with Lucy’s body proportions (see Research Page... Simulation)…both alternative simulations allowed the Lucy model to walk in stable fashion, but, judged by the parameter of mechanical joint power requirements, BHBK walking would have been twice as expensive for her, and (all things being equal) is therefore unlikely…
Using forwards dynamics (‘muscle’-driven) computer simulation models, which can ‘evolve’ their own optimal gaits, however, we can predict metabolic energy costs in different gaits, and allow the simulation to find its own optimal gait, given its proportions and mass distribution. Both for walking and now (when we include serial and parallel-elastic energy stores) running, our predictions for humans fall within 10% of experimentally determined values for equivalent treadmill speeds.
To search for optimum gaits, we are using the genetic algorithm technique, from the field of evolutionary robotics. The robotics part refers to the physics engine and its inputs and outputs, the evolutionary part to the mathematical method of searching for optima, which has been inspired by natural selection.
Our genomes consist of (eg.) body build and muscle activation patterns. First, we form a population of say, 1000 random genomes. These are selected using criteria including for example, the maximum distance covered for a given metabolic energy cost (eg 5000J;) maximum achievable speed, or both. We then pick the best of the 1000 genomes on these criteria and run the simulation again. This is repeated for upwards of 800 generations of the genetic algorithm, which would involve about 800,000 repeats of the simulator, taking about a day on a 42 CPU Beowulf cluster. However, only about 1 in 10 simulations works, the others falling over or otherwise becoming unstable and unable to complete the experiment!
A model of bipedal walking in Australopithecus afarensis AL 288-1 was required to optimize for minimum locomotor costs, and produced values similar to those measured in human children of similar mass/stature by Heglund and Schepens (2003): 7.0 Joules/kg/m for the lowest speeds (c. 0.6 m/s) falling to 5.8 Joules/kg/m at 1 m/s, and rising to 6.2 J/kg/m at the maximum speed achieved. Lucy was probably quite an efficient bipedal walker over short distances. So far, while our human simulations switch to a run at higher speeds, our Lucy model has not so far ‘evolved’ running.
We then used our Australopithecus afarensis model to predict stride length/speed relationships. Slotting in stride lengths of adults, it appears that they were walking at 1.0 m/s or above, well within the range of predicted speeds for an animal of equivalent body size, and moreover, despite A. afarensis’ short stature (c 1-1.13 m), within the range of absolute values for human small-town walking speeds.
But what can the individual footprints tell us? Footprints are made by the application of pressure under the foot, and so may be compared to peak or dynamic records from foot pressure sensors.
Do the 3.6-3.8 mya Laetoli footprints then represent a functionally modern foot, with a fully developed medial arch and eversion/pronation at midstance? Some suggest one of the best-preserved prints may have been produced by a foot used in an inverted/midprone posture, with an abducted hallux (big toe) and no impression of the hallucial metatarsal -- but are these effects artifacts of shading and/or slippage?
Others suggest that this footprint is a good match for a reconstruction of a female Australopithecus afarensis foot skeleton. But how did this hallux form the hollow area to the left?
Since footprints represent the pressure applied to the ground during gait, although slippage and distortion may affect footprint form, we should be able to relate the depth of the print at given points to a given intensity of foot pressure.
This can be done by lofting both to form 3-Dimensional ‘hill’ diagrams.
By studying the type of plot produced in different kinds of human gait, and comparing the pressure plots to laser scans of the footprints made in a substrate during the same stride cycle, we can attempt to match the Laetoli footprints to different types of human gaits.
However, although contour plots reconstructed in 3D from stereopairs give us depth information, they lose textural information which may tell us about slippage, toe flexion etc. And unfortunately since excavation, the Laetoli prints have been damaged by erosion and plant roots over time, so that even modern laser scans do not retain the detail seen in the original images recorded shortly after discovery. Professor Michael Day, who made the original stereopair records shortly after excavation of the site G trails has generously given us access to his unmatched stereo photographs. Our collaborators from the School of Computer Science at the University of Nottingham are using state of the art machine vision techniques to attempt 3D reconstruction without loss of texture detail.
However, we know that the early human ancestor which made the Laetoli prints, whether Australopithecus afarensis or not, would not have had the same body proportions as ourselves, and would not have walked in an identical way as ourselves. We need to use further modelling techniques to create footprints from optimized simulated gaits of models of early hominins. This involves two, parallel, modelling programmes which will eventually be merged.
In the first, we built a 3D skeletal model of a human foot (scaled to a real individual) and drove it to move according to 3D foot motion derived from skin markers by a 6-camera 1000 Hz motion capture system as the individual walked over a pressure-plate. The model was verified by its ability to predict real-world elongation in the soft tissue of the plantar aponeurosis.
The foot model will now be inserted into a lower-body musculoskeletal model, which ‘learns’ the muscle contractions needed to move the foot in that way, and, by a link to a Finite Elements Stress Analysis model, predicts foot pressure – at this point only under the heel – verified by checking predicted deformation in the heel pad against real-world values.
In the other model, we are concentrating on building a finite elements model of the production of footprints by the formation of substrates under foot pressure. We began with a simple 3D spherical indenter model, forming a print in simulated sand under vertical and oblique loads.
We then used a scanned but rigid plantar foot surface to form a 3D print in simulated sand.
A 2D finite elements model of the soft tissues surrounding the foot skeleton during heel strike followed.
The major joints in the foot are now being driven in 2D by motion capture data derived during the stance phase of gait from the marker system used in the first model. The next stage is to use this to indent sand and other substrates, convert the model into 3D and drive it using forces or kinematics derived from our lower body dynamic models of gait in humans and eventually early hominids such as Australopithecus afarensis.
We are therefore progressing steadily towards our goal of being able to predict the kinds of footprints which would be made in the ashes of Laetoli by early hominins such as Australopithecus afarensis moving in different gaits. Then, by reversing the process, indeed reverse-engineering the prints, we hope to be able to determine finally what species made the prints, and how they were walking…………….................or even perhaps, running.