Its latest finding is that being on the point of instability balances straight-line and turning performance in multi-legged robots.
The questions the researchers set themselves was: with all the friction created by multiple legs in contact with the ground, how can a centipede be so agile? And what control strategy is required to get that agility.
A 12 legged robot – a dodecapede – with six two-legged body segments was used (and modelled) for the experiment.
There were no motors between the segments for steering, instead its segments were passively linked by joints and springs.
Its legs were controlled to move so that their tips went straight backwards when in contact with the ground.
As would be expected with this straight contact-point movement, all the segments remained in-line and stay straight and level when the machine walked in a straight line.
Or, at least they did when the inter-segment springs were stiff enough.
But when softer springs were fitted, wiggling oscillations (see photo) spontaneously arose – just like the motion of a real centipede.
Even softer springs meant greater the oscillation and, at a point mathematically described by the researchers as a ‘Hopf bifurcation’, there was a critical spring value which would have the robot on the verge of oscillating motion.
In new work, described in the paper ‘Advantage of straight walk instability in turning maneuver of multilegged locomotion: a robotics approach‘ in Scientific Reports, the team discovered that the robot steers most effectively when it is fitted with springs that keep it close to the point of Hopf bifurcation.
To test steering ability, the front pair of legs were controlled to make an 80° turn to the right, while all the rest were set to walk dumbly forwards – much like a powered trailer.
Fitted with critical springs, the robot reached its new heading in the least time, and fully aligning itself on the new heading most quickly. Stiffer springs meant sluggish turns, and softer springs meant faster turns followed by overshooting the new trajectory and then wobbling around it.
“This study provides clues to unresolved issues of intelligent motor functions of animals, and meaningful insight for biological sciences,” said lead scientist Shinya Aoi.
The Kyoto team also previously showed that the amount of oscillation beyond the bifurcation increases with speed.
This means, the researchers point out, that controlling one factor – inter-segment stiffness – can improve dynamic robot performance without resorting to a precise real-time physics model of all leg motions. At any particular speed there will be a stiffness that will give the most precise steering, and another stiffness that will give the easiest straight line travel.
Combining this with the strategy of only steering with the first pair of legs and letting the rest follow, results in a low computational load within the robot.
Kyoto points out that one of the big threats to robots is falling over, and that more legs reduce the chances of falling. Robots with lots of legs hardly have to think about remaining upright.
In the paper, delightfully, the team also answers the question ‘how many legs does a centipede have? – In real life, from 15 to 191 pairs, or so they have discovered so far.