Abstract:
Transmission dynamics are shown to dominate the stability and performance of
impedance- and torque controlled rotary electro-mechanical systems. The experimental
analysis focuses on planetary, cycloidal, harmonic and cable reducers, but excludes direct-drive,
pneumatic, hydraulic and friction drives. Neither sensors nor actuators with better
resolution nor increased dynamic range can circumvent reduced stability and performance
limitations unless certain hardware criteria can be met Simple transmission models are
proposed to model such effects as (1) transmission stiffness, (2) soft-zones and wind-up,
(3) backlash and lost motion, and (4) stiction, friction and viscous losses. These models are
experimentally verified using six different transmission types most commonly used in robot
designs. Simple lumped-parameter linear/nonlinear models are shown to predict stability
margins and bandwidths at these margins fairly closely. Simple nonlinear lumped- and
fixed-parameter models were unable to properly predict time responses when the torque
signals were of low-frequency and amplitude, underscoring the complexity in modeling the
transmission-internal stick-slip phenomena.
The clear distinction between speed reducers and torque multipliers is theoretically
and experimentally explored. The issue of actuator and sensor colocation is shown to be
extremely important in predicting the reduced bandwidth and stability of torque-controlled
actuator-transmission-load systems. Stiffening transmission behaviors are shown to be of
a conditionally stabilizing nature, while also reducing the dynamic range of impedance- and
torque-servoed systems. System damping, whether active or passive, as well as low-pass
filtering motor-controller signals, are shown to dramatically increase stability without
having any effect on increasing system bandwidth. Transmission soft-zones are proven to
reduce the stability margins of colocated impedance controlled electro-mechanical systems.
None of the standard controller structures explored here were able to noticeably increase the
system bandwidth of the open-loop system, without reducing the overall system
performance.
The different transmissions are tested for system nonidealities and generalizations
drawn on the stability and performance margins of impedance and torque-servoed geared,
cycloidal, planetary, and cable reducers in hard contact with the environment.
Experimental results are furnished which underscore the validity and limitations of the
theoretical modeling approach and comparative transmission analysis, while highlighting
the importance of different physical system parameters necessary for proper transmission
design.
