Biped Howto: Stable locomotion

+++ UPDATE (12-march-2014): Rename title as "Biped Howto".


Along next lines, I will detail the way in which I've implemented the dynamically stable gait, for my robot POPSTAR.

After reading a lot about humanoid periodic gait, stable walking and biped locomotion, I found one that I consider the most clarifying lecture for its understanding. Here you have it.

In the case of the robot POPSTAR, as being built from quite long popsicle sticks, it was impossible to get a statically stable gait. When one leg was in the air, the robot was falling down to the floor. So I had to study the dynamically stable gait. This kind of gait is characterized by several phases in which the robot can be statically unstable. These phases are known as:

  • Double support phase: both feet on the floor, an the robot is statically stable.
  • Pre-swing phase: is a double support phase, but the rear foot is lifting from the floor.
  • Single support phase: one foot on the floor and the other swing through the air. Is statically unstable (if it stops then falls down).
  • Post-swing phase: the swing foot touches the floor again. Is a double support phase.



After several trials, without success, the main problem I had to solve, was that POPSTAR robot was very unstable during the single support phase, and it must swing the foot so fast, to reduce such unstable phase. Even so, it always fall down.

On the other hand, pre-swing and post-swing phases, were also a problem. In both phases, when the rear foot started lifting or landing again, the robot received a little vibration that makes it less stable and it fall down one trial and another.

Finally, I decided to discard those phases and define my own gait.


Once done it, I started to reach a dynamically stable gait, more or less decent (see attached video). In this case, the phases of POPSTAR gait were these:

  • Double support phase: both feet on the floor
  • Single support phase: one foot on the air with the body a bit balanced over the stance foot, cancelling the unstability.

The rear foot followed a triangular path, from the back side to the front. This triangular path was divided into 2 trajectories: from point P0 to P1 and from point P1 to P2.


With this simplified model of the periodic gait, POPSTAR robot, finally get into dynamically stable gait.

That’s an interesting post

That’s an interesting post and brought me closer to understand the whole thing :slight_smile: Please keep working on that.

Hi, I’m working on the same

Hi, I’m working on the same thing using a poppy 3d printed robot platform. I’m using cheap $3 servos and an adapter I made up. I think my problem may be in 12km/cm3 torque specs are A: not powerful enough or B: not the advertised torque rating. Thanks for the diagrams and triangular foot path concept. I’m going to impliment that and see what I can come up with. I’m thinking of using 4 imu’s to help with stability. Granted it’s all theory atm but you gotta start somewhere right.