Why question me? Surely you can see our movements.
Each of us has a characteristic repertoire of movements. You can recognise loved ones just by the way they walk. Actors use different styles of movement to create different characters. Some of these can become iconic, instantly triggering off a complex of ideas, emotions and cultural signifiers. There are basic, gross movements that are common to they way any man walks down a street, but if one of them is Charlie Chaplin twirling an umbrella and the other is John Travolta swinging a paint can, the different personalities are immediately recognisable, and the emotional and cultural connotations are widely different.
It’s important in science fiction drama too. If human actors are to represent alien beings, then finding new styles of movement suitable to the extraterrestrial race in question is essential, if they are not to look simply like a scattering of awkward suburbanites at an unsuccessful fetish party. Wise producers will hire choreographers to work with the actors, giving each species its own palette of movements unique to itself, making each group of aliens seem coherent in itself but distinct from any other.
But what is a style of movement? We can all recognise it, but can we break it down into its elements? Quantify it? Analyse it?
The first project to have a go at pinning down the component elements of dance was commissioned by Louis XIV in the late 17th century. There had been dance treatises before then, elaborate descriptions of how particular dances should be performed (sometimes with stroppy comments about how they should certainly not be performed), but the notation that ballet master Pierre Beauchamp devised for His Majesty was the first to use abstract symbols instead of prose descriptions accompanied by realistic drawings.
This Beauchamp-Feuillet notation, as it became known after Raoul Auger Feuillet popularised it in his many published books of choreography, was an elegant, if initially forbidding, system of swirling lines and sudden angles that represented the motions and transitions of dance just as a set of dots and lines can describe the notes and rhythms of music. It remained in widespread use for a century, before being superseded by a variety of alternative systems.
There are two in wide use today. The Benesh Movement Notation represents body positions on a five-line stave similar to that used in standard musical notation, allowing music and dance notation to be more easily integrated, while Rudolf Laban’s “Labanotation” looks more like geometric abstract art than music, but does have the advantage that it can be used to describe any kind of bodily movement in space and time, not just dance moves.
This idea has been developed further, in Eshkol-Wachman movement notation. Like its predecessors, this breaks down movements into primitive elements, but it uses an elaborate system of three-dimensional polar coordinates to locate these motions in space, with techniques for rotating and translating sequences of movements so that they can be directly compared. This allows the truly invariant characteristics of movements to be calculated.
The applications go far beyond the world of dance. It has been used in a host of animal studies, allowing scientists to establish the movements that are characteristic of particular animals, study how these movements change due to illness or injury, and compare the ways different species of animal move. In one example, Tammy Ivanco and her colleagues from the University of Lethbridge, Canada, used Eshkol-Wachman notation to quantify the different ways that rats and opossums reach for food, and were able to relate the more complex movements of the rats’ hands and arms to their relatively more elaborate brains and nervous systems.
It may even prove useful in studying the human brain. Autism is not generally diagnosed until a child is around three years old, while Asperger’s Syndrome is diagnosed much later – typically around the age of six or seven, but it can remain undiagnosed into the teenage years. Osnat Teitelbaum and her colleagues at the University of Florida analysed video recordings of infants moving about, and by using the Eshkol-Wachman system were able to determine certain movement styles that were characteristic of children who would later be diagnosed with autism or Asperger’s Syndrome. These were things like asymmetric crawling, where the infant would not crawl in the efficient manner of most babies, moving diagonally opposite limbs together, but would instead move in clumsier ways, such as with one foot stepping while another crawls, or a particular way of falling forward or back from a sitting position without using the reflexive motions of the arms that neurotypical infants would protect themselves with. This work led them to develop a simple motion-based test for autism and Asperger’s Syndrome in infants, whereby the child is held and the waist and slowly tilted from side to side. If the infant does not manage to keep their head vertical, an autistic spectrum disorder may be present.
A much simpler form of notation was devised recently by Amy LaViers, an engineering postgrad at the Georgia Institute of Technology. (That’s Georgia the US state, not Georgia the former Soviet republic.) Eschewing the complexity and power of the Eshkol-Wachman notation, LaVier’s system represents two legs, each of which can adopt one of ten different poses. The sequence of poses, and the transitions between them, describe the dance.
These ten discrete states are not chosen arbitrarily. Ballet dancers perform their warm-up exercises at the barre, a handrail that they hold on to for stability as they exercise each leg in turn. The ten barre exercises are the building blocks of ballet, and it is these movements that are captured in LaVier’s finite state automaton, a computer program that moves through these different poses to create sequences of dance.
There are constraints on the movements the automaton can perform. Some of these are physical – it cannot hover with both legs off the ground like some Jedi Cossack – but others are aesthetic. Specific mathematical constraints define the style and content of the dance, and as the automaton improvises within these constraints the audience perceives the character of its motion.
The aim of this work is not to create a ballet-dancing robot. Rather, it is to find ways to make robots move with particular styles and qualities. Non-verbal communication is expected to become an important element of the human-machine interface, as machines become more mobile and autonomous. A Predator drone may have no need to appear friendly (though for PR purposes I can imagine one of its successors might), but as robots increasingly interact with humans in non-lethal contexts, their body language may be the critical factor in putting people at their ease.
In this way, the robot engineers face the same sort of challenge as a choreographer on a science fiction show. They each have to define characteristic styles of movement that their performers – actors or robots – can work within, generating arbitrary sequences of movement that remain within strict aesthetic constraints. The difference is that the choreographer wants to make the actors seem as inhuman as possible, moving with a sense of the strange and uncanny, while the engineer wants the robots to seem as human, friendly and familiar as an automaton of motors and software can be.