|
| |
|
|
|
|
||||
| Home Help Feedback Subscriptions Archive Search Table of Contents | ||||
Inside JEB |
MR AVERAGE BLAZES LUCY'S TRAIL
kathryn{at}biologists.com
|
Sometime in the past, our ape-like ancestors descended from the trees, and took to life on two feet; but `when' and `how quickly' we took to a terrestrial lifestyle is hotly debated. Bill Sellers is keen to know how ancient humans might have walked during this momentous time in human evolution. However, with little physical evidence of our ancestors' build, Sellers knew he'd have to develop a method that could successfully reproduce a modern-day biped before attempting to recreate our ancestors' early steps. Teaming up with Louise Dennis, an expert in computational artificial intelligence, and Robin Compton who already had experience of simulating early human movements, Sellers began testing kinematic calculations to see if he could first produce a walk with Mr Average (p. 1127), before heading back several million years to simulate `Lucy'.
Most kinematic simulations are based on a detailed understanding of the animal's physical build and stride pattern. But Sellers explains that there is only one skeleton for an ancient human living at the time we might have made our final descent from the trees, and even though Lucy is largely intact, it is almost impossible to predict how she moved based on her build alone. But knowing that modern human walking is optimised to give us the maximum range at the least energetic cost, Sellers decided to ask the question the other way round; could he simulate Lucy's motion by simply asking her to walk as far as possible on a given amount of energy without falling over? Maybe he could, but first he'd have to prove that his simulations worked by testing the method on modern humans first.
Using the genetic algorithm, a powerful computational method, Sellers began running millions of tests to see how far Mr Average could walk on 5000 J. First he randomly chose starting conditions, but the computated-person either ran out of energy within a few steps, or stumbled to a halt.
Sellers knew he'd have to search wider to find an efficient gait, so he simulated three million starting conditions, whittling the enormous field down to 100 that successfully walked for 3 m, before setting them off to see who had the greatest stamina. Sellers was delighted when the results produced some efficient, but unconventional walking styles. He explains that this could be very useful, as simulations of bipedal dinosaurs have always assumed that they basically walked like us. The flexibility of the simulations to discover new gaits could help us to discover how other bipeds once walked the earth.
But the most realistic human simulations came out when Sellers gave them a helping hand. He explains that he spent a morning sitting at a computer, estimating the best set of starting conditions to set his model man striding out across his computer screen. The results are very convincing, and even though the walk is a little stiff legged, it's as efficient as any flesh and blood human, covering 25 m before running out of energy.
Having convinced himself that the genetic algorithm reproduces a modern human's gait well, Sellers has returned to the past, and is beginning to simulate how Lucy might have walked 6 million years ago. And he adds that so far `Lucy's walking is coming along well'.
References
Sellers, W. I., Dennis, L. A. and Crompton, R. H.
(2003). Predicting the metabolic energy costs of bipedalism using
evolutionary robotics. J. Exp. Biol.
206,1127
-1136.
Related articles in JEB:
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||
