Abstract:
Friction is the tangential reaction force between two surfaces in contact. It is present in every
mechanical system. Various models exist in literature that captures the effect of friction. Some
are static while others are dynamic. Dynamic models have the advantage over static models in
their ability to capture the position and velocity dependent phenomenon and provide models that
closely correspond to real life systems. Friction deteriorates performance of physical systems.
Low velocity motion can’t be achieved without first determining the various disturbances such as
cogging torque, friction and external disturbances. Friction is the main cause of performance
degradation. By determining how friction is affecting a particular system and properly
compensating it can help in improving the regulation/tracking and effectively reducing the error.
Friction can’t be compensated until it is known. For this purpose two friction models, one from
the static and the other from the dynamic set of models have been chosen. The parameters of
both models have been estimated. The estimation is based on off-line techniques. The estimated
parameters have been verified by applying model validation procedure. The estimated
parameters are plugged into a friction model that forms the core of the friction compensator. In
this thesis, Coulomb and LuGre friction models are considered. The Coulomb only consist of a
single parameter whereas the LuGre constitutes of static and dynamic parameters. A single link
robotic manipulator is used as a test-bed. Simulation and experimental studies have been
conducted for three cases, namely; no compensator present, Coulomb model and LuGre model
present in the control methodology. Separate experiments are conducted for the estimation of
parameters of LuGre friction model on hardware. Friction compensation is then achieved using
these parameters. Performance is assessed by the tracking errors of the various schemes and the
control effort required to track a particular trajectory. Further analysis is done by the comparison
of phase portraits and additionally, error phase portraits have also been plotted.
