The Tricky Transition from Walking to Running

Biomechanical engineers say we can switch from walking to running without increasing our energy use.

The way humans walk and run has puzzled robotics engineers for many years. This process occurs with little conscious or subconscious control. It’s almost as if our legs know how to do it.

Recently, however, biomechanics experts have begun to tease apart the processes involved and it does indeed look as if legs walk and run with only the highest level of intervention from the nervous system.

That implies that with the right kind of design, robot legs should work just as well with little central control.

This goes against many decades of thinking that every movement of every part of a robot needs to be calculated in advance by a central processing unit and then monitored and carefully executed.

Every attempt to build robots like this has run into a problem: there’s no way to carry enough computing power, sensors and actuators to do this.

Instead, biomechanics are now focusing on a new approach based on an idea called passive dynamics. This is essentially design for movement but with little or no central control.

A good example is the difference between a high performance jet, which requires fly-by-wire controls and powerful computers to keep it flying nose first, and a dart, which by design always fly nose first, however it is thrown.

Biomechanics say a similar principle of passive dynamics must control our gait but the difficulty is in finding out how it works.

One approach is to think of the leg as an inverted pendulum during walking and as a spring loaded inverted pendulum during running. In this model, the movement is controlled by factors such as the timing of each step and leg lift and the angle the leg makes with the ground when it lands. Various groups have studied this model and shown it to be a reasonable abstraction of the way we walk (and the way many animals walk too).

However, one problem is to work out how the switch from walking to running occurs, while ensuring that the gait is stable throughout. It’s no good if the transition triggers an instability that causes the robot (or human) to fall over.

Today, Harold Roberto Martinez Salazar and Juan Pablo Carbajal at the University of Zurich reveal their work on this problem. These guys have attempted to characterise the entire ‘space’ of movement for both walking and running and then worked out how best to make the transition.

They’ve found that walking and running spaces both contain regions that are stable and unstable. Obviously it’s a good idea to avoid the unstable regions when walking or running.

However, their new insight is that the regions of instability can be used to make the switch from one gait to another.

That makes sense. If your walking is entirely stable, how do you make the switch to running? Something has to change. And by using a region of instability, it’s possible to catapult from the walking space into the running space.

And what’s interesting is that this transition can be made at a constant energy level. So no additional power is needed to make the switch.

Of course, the idea is not that a robot would calculate how to make the transition across the movement space. Instead, roboticists would have to design the robot legs so that they can easily access the required areas of instability and stability simply by changing parameters such as the angle and timing at which a leg hits the ground.

That’s the theory anyway. The Zurich team will be only too aware that actually designing and building legs that do the trick will be much harder.

Ref: arxiv.org/abs/1108.4432: Exploiting the Passive Dynamics of a Compliant Leg to Develop Gait Transitions