We were once toddlers before. We’ve gotten drunk as recent as a few days ago. We maybe got too sore from working out. And we’re not always perfectly balanced. We know how awkward it can be sometimes to just… walk. It’s OK! It’s not always as easy as it looks! Look, even computer simulations tasked to figure out how to walk sometimes fall face down on the floor.
In a video posted by John Goatstream, a computer simulation made by Thomas Geijtenbeek, Michiel van de Panne and Frank van der Stappen shows how muscle-based bipeds could walk in different simulations. They can tweak the simulation to walk under certain types of movement, gravity, obstacles, body types and so forth. The video describes it as such:
All actuation forces are the result of 3D simulated muscles, and a model of neural delay is included for all feedback paths. As a result, our controllers generate torque patterns that incorporate biomechanical constraints. The synthesized controllers find different gaits based on target speed, can cope with uneven terrain and external perturbations, and can steer to target directions.
It’s like letting a computer figure out the best way to walk for different two-legged animals. Pretty cool. Really cool to see that some models struggle with the basic concept of walking too.