Premise

Our goal was to develop a back-mounted camera stabilization platform for use on rugged terrain, specifically for park rangers who might be interested in documenting trail condition, though it could also be used for independent film crews.

Controller   desired

Needed a controller to compensate for the high-frequency disturbances induced from waking and the low-frequency motion from walking up and down.

Approach

  • Developed CAD model
  • Developed Inverse Kinematics model
  • Generated sinusoidal walking input with noise as sensor error (5cm amplitude with a frequency of 10 radians/second)
  • Developed low frequency and high frequency controllers
  • Optimized model for realistic torque outputs

Conclusion

  • Controller compensates for disturbances from gait pattern
  • Current motors not sufficient for all scenarios
  • Low-pass filters do not always work for sensors​

Future Work
  • Applying a bayseian model and a predictive controller to adapt to more than just flat terrain
Picture
Proposed design/mounting strategy for such a system
Our project was somewhat inspired by this video.