The development of legged robots has come a long way, but concerning speed and energy efficiency there are still crucial steps to be taken. For inspiration we can take a look at nature, where quadrupedal runners far outperform anything mankind has produced thus far.

Morphological computation

Attempts to make high-performance quadrupeds are often based on a mechanical design that “looks like” an animal such as the leopard. From there, controllers are optimized for speed, energy or stability—but this will not yield the desired performance. Embodied (artificial) intelligence or morphological computation is the idea that system behavior is the result of the geometrical interconnection of elements such as springs and inertias. This embodied AI leads to inherent behavior, in this project a (fast, efficient) running motion.

Interconnections based on power

Port-Hamiltonian Systems (PHS) theory models physical systems as the (geometrical) interconnection of energy storage and dissipation elements, which not only gives direct insight to the power flows, but makes PHS theory by its nature the perfect candidate to study morphological computation.

Project goals

In this research project we will design a quadrupedal runner that has inherent running behavior, caused by proper interconnection of storage elements: then a control system only needs to supply energy to keep this motion going and stabilize it when necessary, resulting in a fast, energy-efficient running quadruped.

As inspiration for the design we will look at nature’s best runners—not by copying the mechanical design, but by identifying them as a PHS and mimic the energetic behavior, copying that embodied AI that leads to running motions. 

Click here to see a movie

Project lead

no picture available Stefano Stramigioli