Exploitation of augmented passive dynamics for bipedal locomotion


Exotendons can be used as a passive component to decrease the energy needed to accomplish movement. Therefore they can help people with mobility issues or augment the autonomy of a robot or a powered armor by providing some of the power.

This semester project is based on Bogert’s paper “Exotendon for assistance of human locomotion” which provide a statical analysis of the use of exotendons throughout several configurations. The goal is to study the impact of exotendons systems on the human gait in a dynamic environment with the help of a simulation program and to compare the result obtained in the dynamic simulation with those obtained by a static analysis. Additionally, we should search for new exotendons configurations and new passive components to add to the model.

The model we use in our simulation.

 We began by implementing the same configurations than in the paper, then we added our own and optimized all of them using the Particle Swarm Optimization. We completed the simulation by modifying the contact model to simulate a damper in order to diminish the shock when the foot hits the ground. When we had everything correctly running and optimized, we proceeded to a careful analysis of all the parameters that acts on the residual torques and power, that is, the torques and power that need to be provided by the servomotors to produce the movement. To do the analysis, we helped ourselves with a lot of graphs that represent the status of the gait throughout an entire gait cycle.


Report and Presentation


The first video shows an issue that we have when initializing the gait: since the humanoid goes from an immobile phase to a running phase with no acceleration phase between the two, the collisions with the ground are more sharp than they should be, resulting in an instable beginning of the walk, even though we tuned several parameters to minimize this initial instability. The gait however quickly corrects itself and there isn’t any issue afterwards anymore. 

The second video shows the gait of our model. Note that the foot slightly sinks into the ground, this is due to the customized contact model that we used to simulate the softness of the contact when the foot lands.
Note that in both videos, the walk appears a bit slower than it should be due to performance issue caused by the recording.