Useful resource-sharing boosts robotic resilience – Robohub

The Mori3 modular origami robotic. Picture credit score: EPFL. Reproduced beneath CC-BY-SA.

By Celia Luterbacher

If the objective of a robotic is to carry out a operate, then minimizing the opportunity of failure is a prime precedence with regards to robotic design. However this minimization is at odds with the robotic raison d’être: techniques with a number of models, or brokers, can carry out extra numerous capabilities, however additionally they have extra completely different components that may doubtlessly fail.

Researchers led by Jamie Paik, head of the Reconfigurable Robotics Laboratory (RRL) in EPFL’s Faculty of Engineering, haven’t solely circumvented this drawback, however flipped it: they’ve designed a modular robotic that really lowers its odds of failure by sharing assets amongst its particular person brokers.

“For the primary time, we’ve got discovered a method to reverse the pattern of accelerating odds of failure with rising operate,” Paik explains. “We introduce native useful resource sharing as a brand new paradigm in robotics, lowering the failure price with a bigger variety of modules.”

In a paper printed in Science Robotics, the staff confirmed how exploiting redundant assets and sharing them domestically enabled a modular origami robotic to efficiently navigate a fancy terrain, even when one module was fully disadvantaged of energy, sensing, and wi-fi communication.

Sharing is caring

The RRL staff took inspiration for his or her innovation from nature, the place the issue of failure is commonly solved collectively. Birds share native sensing info via flocking conduct, some bushes talk threats to neighbors utilizing airborne indicators, and cells repeatedly transport vitamins throughout their membranes in order that the loss of life of any particular person doesn’t considerably impression the general organism.

Modular robots, that are composed of a number of models that connect with kind a whole system, are analogous to multicellular or collective organisms, however till now, their design has been a supply of vulnerability: the failure of 1 module usually disables some, if not all, of the robotic’s skill to carry out duties. Some modular robots get round this drawback with built-in backup assets or self-reconfiguration talents, however these approaches often don’t fully restore performance.

For his or her examine, the RRL staff used one thing referred to as hyper-redundancy: the sharing of all important energy, communication, and sensing assets throughout all modules, with none change to the robotic’s bodily construction.

“We discovered that sharing only one or two assets was not sufficient: if every useful resource had an equal likelihood of failure, system reliability would proceed to drop with an rising variety of brokers. However when all assets have been shared, this this pattern was reversed,” Paik says.

In a locomotion activity experiment with the Mori3 robotic, which consists of 4 triangular modules, the staff experimented with chopping battery energy, wi-fi communication, and sensing to the central module. Usually, this ‘lifeless’ central module would block the articulation and motion of the opposite three, however due to hyper-redundancy, the neighboring modules totally compensated for its lack of assets. This allowed the Mori3 to efficiently ‘stroll’ towards a barrier and contort itself successfully to cross beneath it.

“Primarily, our methodology allowed us to ‘revive’ a lifeless module in a collective and produce it again to full performance. Our native resource-sharing framework due to this fact has the potential to help extremely adaptive robots that may function with unprecedented reliability, lastly resolving the reliability-adaptability battle,” summarizes RRL researcher and first writer Kevin Holdcroft.

The researchers say that future work may deal with making use of their useful resource sharing framework to extra complicated techniques with rising numbers of brokers. Specifically, the identical idea may very well be prolonged to robotic swarms, with {hardware} variations that permit swarm members to dock to one another for power and knowledge switch.

References

Scalable robotic collective resilience by sharing assets, Holdcroft, Okay., Bolotnikova, A., Monforte, A.J., and Paik, J., Science Robotics (2026).