It’s taken a while, but now the Unibot has some stabilizing capabilities in the lateral (side to side) direction as well. In the movie it can be seen that the robot does not reject disturbances very well yet. This still needs some improvement.

 

In the experiments the lateral motion was decoupled from the transversal dynamics by mounting the robot in a frame that only allows lateral motion.

 

System identification was carried out with a hanging robot. In such a way the (hanging) dynamics can be identified in open loop (no control needed). When the hanging dynamics has been found, the standing, unstable model can be obtained by rotating the dominant complex poles and zeros of the system by 90 degrees. This will be made more clear in my thesis.

 

The controller that is used in the experiments was designed using H-infinity loop shaping. For this part I have to acknowledge Raymond de Callafon for his support.

 

The next step would be to combine the transversal and lateral motion and controllers. However, due to the limited capabilities of the lateral controller, it is probably too early to try this.