Watch human cross-country runners and you can't fail to be impressed by the way they adapt to different surfaces. Andrew Spence explains that human runners achieve this by a subtle trick. ‘If you look at the force patterns acting on human runners they look like what you would predict from a pogo stick, which is a point mass with an elastic spring leg coming off the bottom that bounces whenever a foot hits the ground,’ explains Spence. When humans run over soft surfaces, they effectively stiffen the virtual ‘pogo stick’. But humans are the odd-ones-out in the animal kingdom. Most animals run around on four or more legs, so how do animals with multiple legs adjust to running on soft surfaces (p. )?
Spence became intrigued by this problem when he was working in Robert Full's Poly-PEDAL lab at U. C. Berkeley. Talking to his colleague, Shai Revzen, they realised that they could use the minute accelerometers that Spence was building to see if animals also adjust the stiffness of their virtual ‘pogo stick’ legs. According to Spence, animals also move as if they are bouncing on virtual pogo sticks when running on hard surfaces, no matter how many pairs of legs they have. So how do they cope when the going gets soft?
Gluing one of the tiny accelerometers to a cockroach's back, the duo sent it scuttling off across a sheet of latex. The insect sunk into the soft rubber surface, like we sink into mud, but kept on going at the same speed. And when the duo looked at the wavy acceleration plots generated by the insect, ‘we got beautiful wiggly lines,’ remembers Spence. But the accelerometer could not give the duo all the information they needed. The team also had to measure the insect's position, orientation and velocity; and the only way they could get that information was by attaching a light balsawood cross to the accelerometer backpack and tracking the insect's movements with video.
After a year of development, the system was ready for Chris Mullens to collect data as the cockroaches scurried across the latex sheet. Having measured the accelerations experienced by the floundering insects, the team had to see if they could reproduce the forces that produced the real cockroaches' accelerations in a computational cockroach.
Building a computer model of a cyber-cockroach with its six legs replaced by a pogo stick that bounced every time three of the insect's feet hit the ground (two pogo stick bounces per stride cycle), the team found that when the cyber-roach ran on latex it didn't look right. The simple pogo stick model could not explain how the insects ran on, and sunk into, soft ground.
The team needed a more representative model, so Justin Seipel, an applied mathematician in Full's lab, suggested adding a clock that drives the pogo sticks with a motor at the hip. Could this model reproduce the forces exerted on the cockroaches as they scuttled across the latex?
Amazingly it did. The cyber-roach kept moving at the same speed as it hit the virtual latex, but as it sunk into the soft surface, the trio of feet moving through the air hit the ground sooner than if the insect was on a hard surface. Instead of reducing the forces on the body's centre of mass by stiffening each virtual pogo stick leg – like human runners – Spence suspects that the insect's inefficient posture, as its feet hit the ground too soon, could reduce the forces acting on the insect's body without having to change the stiffness of its virtual pogo stick legs.
Spence explains that this simplifies control. Instead of sending nervous system signals to stiffen muscles, the cockroach may just be able to continue sending the same control signals and take advantage of the change in posture. ‘By putting the intelligence in the mechanics you can reduce the amount of computation you have to do. It's a winner all round,’ says Spence.
