Original source (on modern site)
Robots have certainly come a long way since the early days of robotics where they were primarily confined to industrial applications like welding and simple operations on assembly lines. Today, these intelligent machines have permeated nearly every facet of our lives, from domestic settings with vacuum-cleaning robots to complex autonomous vehicles navigating our roadways. Despite these advances, designing robots that can efficiently interact with real-world environments has proven to be challenging, and even modern systems often leave much to be desired. These challenges are increasingly leading engineers to look to nature for inspiration. Nature offers a wealth of strategies that provide innovative solutions for navigating diverse terrains, detecting and responding to environmental stimuli, and solving complex problems. HAMR-Jr was inspired by the cockroach (📷: Kaushik Jayaram) But as we attempt to copy these designs from nature, it is abundantly clear that even our best performing robots lag far behind their biological counterparts. A team led by researchers at the University of Washington wanted to understand why that is. After all, we have some incredibly advanced energy storage and actuation systems, so why can they not beat nature at its own game? As a first step in unraveling this mystery, they took a close look at the factors that allow animals to run faster than the robots designed to mimic them. The mystery only grew deeper as they delved into the details of the five critical subsystems that are most prominently involved in running, namely the power, frame, actuation, sensing, and control subsystems. As they examined each component, they found that the artificial system offered better performance. Consider the power subsystem, for example. A good lithium-ion battery can supply up to 10 kilowatts of power per kilogram. That is enough to make a cheetah jealous — animal tissue can only generate about 10 percent as much power. Similar findings were uncovered when looking at actuators. Muscle tissue cannot come close to the torque that can be produced by a powerful motor. After puzzling over how these findings can be squared with the observation that animals can run laps around these technically more powerful robots, the researchers came to the conclusion that the present problem arises from a lack of integration between the subsystems. The separation that we see between robotic components simply does not exist in the natural world, where systems are far more integrated. The team points to the way that the leg muscles, for example, are actuators that propel the body, but they are not only actuators. They also have the ability to generate power by breaking down fats and sugars, and can sense their environment by using neurons that are embedded within them. The team proposed that rather than dividing subsystems into individual units as we do today, we should instead consider breaking the systems down into what they call "functional subunits." This would involve changes like moving power sources and actuators into the same units, such that they operate more like biological tissue. Solving entire system-level problems such as this one is very challenging, so it would require an entirely new way of looking at robot design and a new set of tools to support their development. But at this time, this proposal has not been backed up by a robot design that actually outperforms animals, so further validations will be needed to determine if this approach is the best path forward.