Lessons from a Decade of Walking Robots
A retrospective over ten years of garage-built walkers: which mechanisms survived contact with reality, and which control assumptions did not.
Build along
Timestamps jump straight to the matching moment in the video.
Before control theory, get a joint that holds position under load and reports its own angle. Most early failures are mechanical, not algorithmic.
A walking pattern is only useful once the robot can keep its centre of mass inside the support polygon at rest.
Printed parts are excellent for iteration and poor for high-cycle load paths. Put metal where the torque is.
Actuator sizing, fall recovery and the habit of testing one change at a time carried straight over from quadrupeds.
James Bruton
@JamesBruton
Taught by this creator. Follow to see new tutorials in your feed.