Rafik Hariri philanthropic and developmental contributions are countless. The most remarkable being the multifaceted support to educate more than 36,000 Lebanese university students within Lebanon, and beyond.
You are here
ROBUST ADAPTIVE CONTROL OF FLEXIBLE JOINT ROBOTS
Primary tabs
Fouad T. MRAD
|
Univ. |
Purdue |
Spec. |
Electrical Engineering |
Deg. |
Year |
Pages |
|
Ph.D. |
1990 |
137 |
During the 1980's, joint stiffness of industrial robots was experimentally observed and described by constant torsional springs. It was concluded that neglecting the joint flexibility in control strategies limits the robot's ability to perform high speed and high precision operations. Robots were introduced into industrial environments to increase production and to lower costs. The need for reprogramming with different loads and tasks, wasted valuable time, therefore adaptive instead of fixed control laws were desirable.
The desired actuator trajectory in a flexible joint robot is dependent not only on the desired kinematic trajectory of the link but also on the link dynamics. Unfortunately, link dynamic parameters are unknown in most cases, as a result the desired actuator trajectory is also unknown. To overcome this difficulty, a number of control schemes require the feedback of link acceleration and link jerk. In this thesis we describe three control schemes for flexible joint robots which do not use link jerk or acceleration. One of the controllers is suitable for trajectory tracking when the robot parameters are known in advance. The other two control laws are derived from candidate Lyapunov functions, which resemble the energy of the arm deviating from the desired trajectory. Trajectory tracking and adaptation of robot arm parameters are possible with two of the controllers described in this thesis. Our control schemes do not require the numerical differentiation of the velocity signal, or the inversion of the inertial matrices.
There are many directions in which our work can be extended in the future. Some of these possible extensions are now outlined.
i) A natural future extension is the insertion of boundary layers to prevent the switching structure of our two discontinuous controllers.
ii) In addition to model the joint flexibility and actuator dynamics, link flexibility can be added to the modeling process.
iii) Since most industrial robots use feedback sensors mounted on the actuators, the possibility of observing the link angular position and velocity is an important future research problem.
iv) Finally, experimental work can be done to demonstrate the practicality of our schemes.
Applications of modeling and controlling flexible joint robots can be extended to manipulators constrained motion, optimal control, coordinated multiple robots, etc.







