Next: Dynamic ModellingIdentification Up: Identification and Control Previous: Torque Control of

Control of Impedance of IREQ's Dextrous Arm

Authors: [tex2html_wrap4342]M. Doyon, V. Hayward

Investigator username: hayward

Category: robotics

Subcategory: identification and control

For a manipulator to interact predictably with its environment, it is possible to program its visco-elastic properties by feedback control, so they assume simple forms such as strictly diagonal stiffness and damping matrices. The sensing and actuation system of IREQ's GRLA (General Large Robot Arm, a.k.a the ``Gorilla'' arm, Figure 30) SARCOS manipulator lends itself to decentralized control strategies to acheive this goal. However these strategies cannot in general guarantee the achievement of simple forms for the visco-elastic properties, even when its human-like kinematic redundancy is put to contribution to select the most favorable arm postures. We are proposing a generalization of the basic decentralized control scheme which can achieve any desired form of the stiffness and damping matrices within the limits of the actuators operational capacities.

cliquesptitle Applied System Identification Authors: [tex2html_wrap4344]I.W. Hunter, S. Lafontaine

Investigator username: hunter

Category: robotics

Subcategory: identification and control

System identification techniques are developed in order to characterize robotic systems (e.g. robot actuators), so that the resulting models may be used to formulate near-optimal controllers. A main development is the use of Chebychev polynomials for representing the system as well as its inputs and outputs.


baron@cim.mcgill.ca