Aspects of vision-based sensing and control for space robots.

Abstract

The objective of this study is to advance the state-of-the-art in the control of space manipulators. In the microgravity environment of space, large robots like the Space Station Remote Manipulator System are able to manipulate enormous payloads, with masses several orders of magnitude larger than the mass of the manipulator itself. Resulting low-frequency structural oscillations give rise to dynamic stability problem when the robot is controlled by a vision system measuring the payload motion. This problem is not unique to space robotics applications but is generic for the control of structurally flexible systems when the motion is measured at a distance from the actuator controlling the motion. This study develops analytical criteria for the selection of the number, type and location of suitable sensors for robust control of mechanical systems with flexible bodies. Sensors meeting these criteria are called "hyperstability sensors". A dynamic sensor data fusion approach is developed to integrate additional "performance sensors" such as vision-based sensors, addressing the problem of using non-collocated sensors and actuators in the active control of flexible structures. An extended hyperstability concept is developed to enable robust control of complex systems with conventional or intelligent control systems. The model-free methodology is applicable to the broad class of non-linear and time-variant system governed by Hamilton's Principle which encompasses most mechanical systems. The new methodology is applied to a flexible link robot and experimentally validated. The tests involve the measurement and control of the end-point motion of the flexible link using a vision system and real-time image processing. The methodology is shown to be effective in the experiment, enabling robust control as expected from the theory, and relatively simple to implement. Simulations complementing the hardware tests extend the scope of the evaluation. A comparison with a Kalman-filter-based control design confirms the superior robustness properties of the proposed approach.