Back in the late 1980s, a Canadian engineer called Tad McGeer built a remarkable pair of mechanical legs that were unpowered, had no actuators, no sensors and no computer control. But set them in motion down a slight incline and they started to walk with the lazy rolling gait of a gunslinging cowboy.
By contrast, robotic legs are packed with sensors to monitor the position of each joint, computer processors to plan the trajectory of every movement and actuators to push the limbs into position. All this, of course, is power-hungry work.
McGeer’s breakthrough raised the prospect of an entirely new approach to robotics in which much of the intelligence required to control the machine is built into the design. He called this new approach passive dynamics and the hope was that it would make robots almost as capable and energy efficient as humans.
But things haven’t quite worked out like that. More than 30 years later, engineers have certainly made progress, but robots are still power-hungry machines packed with sensors, actuators and processors. More efficient machines are certainly possible so an important question is how little electronic help can a robotic walker get away with? Or in other words, what is the simplest way to achieve walking?
Now we get an answer thanks to the work of James Kyle at Carnegie Mellon University in Pittsburgh and colleagues who have built a walking robot called Mugatu that consists of just one motor and two rigid legs.
Kyle and co say their design is the first bipedal walking robot to contain of this simplicity that can start and stop under its own steam, walk in a straight line or steer. “Despite its morphological simplicity and single actuator, it is also able to steer left and right and smoothly vary its turning radius,” they say.
The work pares down the machinery required for robotic walking to the bare minimum. And since Mugatu doesn’t even have knee joints, its simplicity paves the way for other simpler and even smaller walking devices.
The act of walking is deceptively complex. A walking robot must lift each foot and move it forward while maintaining its balance and being able to steer.
Kyle and co say that this usually requires several different motors to coordinate all these tasks. “We show that careful mechanical design subject to several design rules enables one motor to conduct all these tasks at once,” say the team.
The design and curvature of Mugatu’s feet turn out to be crucial because they allow the robot to roll back and forth and also from side to side as it walks. Kyle and co have discovered a number of design rules that ensure the robot can walk despite its mechanically simple setup.
For example, one design rule is that the robot’s center of gravity must be below the center of curvature for the feet. This ensures that when the robot is off balance, it rolls back to an upright position.
A second rule is that when the robot stands upright with its feet together, it must lean backwards. This ensures that when the hip rotates to move a leg forward, the leg lifts upward rather than jamming against the ground.
With its leg up, the robot rolls forward, leaning towards the lifted leg. When this hits the ground, the other leg lifts and moves forward causing the robot to lean the other way. And so on.
When walking, the curvature of the feet determines the frequency of this rolling gait. So matching the action of the motor to this frequency causes the robot to walk in a straight line. Changing this frequency causes the robot to veer left or right.
The team go on to discuss other factors such as the robot’s inertia -- each leg contains a battery which increases its mass, although Mugatu is powered by just one of them.
Mugatu turns out to be an efficient walker compared to other robots of a similar size. Indeed, the faster it walks the more efficient it becomes. This occurs with a lower step frequency but longer strides.
That’s interesting work that takes the principle of passive dynamics to a new level. There is much further to go however. The team point out that even though Mugatu outperforms other robots it is no match for the efficiency that animals achieve.
But it is still early days for the team. “Future design optimization studies could investigate if efficiency could be improved,” they say.
In the meantime, engineers can surely learn from Kyle and co’s use of passive dynamics to make robots better, one step at a time.
Ref: The Simplest Walking Robot: A bipedal robot with one actuator and two rigid bodies : arxiv.org/abs/2308.08401