Direct Model Reference Adaptive Control of a Flexible Robotic Manipulator

Quick, precise control of a flexible manipulator in a space environment is essential for future Space Station repair and satellite servicing. Numerous control algorithms have proven successful in controlling rigid manipulators with colocated sensors and actuators; however, few have been tested on a flexible manipulator with noncolocated sensors and actuators. In this thesis, a model reference adaptive control (MRAC) scheme based on command generator tracker theory is designed for a flexible manipulator. Quicker, more precise tracking results are expected over nonadaptive control laws for this MRAC approach. Equations of motion in modal coordinates are derived for a single-link, flexible manipulator with an actuator at the pinned-end and a sensor at the free end. An MRAC is designed with the objective of controlling the torquing actuator so that the tip position follows a trajectory that is prescribed by the reference model. An appealing feature of this direct MRAC law is that it allows the reference model to have fewer states than the plant itself. Direct adaptive control also adjusts the controller parameters directly with knowledge of only the plant output and input signals. No a priori knowledge of the plant is necessary. Simulations are performed to test both nonadaptive and adaptive model reference control on the flexible manipulator model. Although nonadaptive control gives satisfactory tracking results, the adaptive control does not due to the inability of the noncolocated system to satisfy a necessary positive realness condition. When the sensor and actuator are nearly colocated excellent tracking results are achieved.