Machine Vision Measurement of Compliant Five Bar Mechanism
Disciplines
Other Computer Engineering | Robotics
Abstract (300 words maximum)
Despite the numerous advantages of compliant and flexi-
ble mechanisms, since the flexible members deflect more and it’s
hard to predict their deformation under complex loading, the dy-
namical modeling of mechanisms incorporating compliant links
and flexures is much more complicated than rigid body model-
ing. Although finite element modeling provides accurate analy-
sis deeming the material nonlinearities, an accurate mathemat-
ical model is still required for control purposes. In this study,
we present the design and reinforcement learning-based model-
free trajectory control of a 2 DOF compliant, serial, and closed-
chain compliant mechanism incorporating large deforming flex-
ure hinges.
For the target planar closed chain fully compliant
mechanism, we implemented machine vision measurement to precisely read and record the planar motion of the selected points. Once the input and output relation is known, it'd allow us to model the compliant mechanism using reinforcement learning.
Academic department under which the project should be listed
SPCEET - Mechanical Engineering
Primary Investigator (PI) Name
Ayse Tekes
Machine Vision Measurement of Compliant Five Bar Mechanism
Despite the numerous advantages of compliant and flexi-
ble mechanisms, since the flexible members deflect more and it’s
hard to predict their deformation under complex loading, the dy-
namical modeling of mechanisms incorporating compliant links
and flexures is much more complicated than rigid body model-
ing. Although finite element modeling provides accurate analy-
sis deeming the material nonlinearities, an accurate mathemat-
ical model is still required for control purposes. In this study,
we present the design and reinforcement learning-based model-
free trajectory control of a 2 DOF compliant, serial, and closed-
chain compliant mechanism incorporating large deforming flex-
ure hinges.
For the target planar closed chain fully compliant
mechanism, we implemented machine vision measurement to precisely read and record the planar motion of the selected points. Once the input and output relation is known, it'd allow us to model the compliant mechanism using reinforcement learning.