Real-Time Configuration Control System for Redundant Manipulators and Analysis of Avoidance Space
Tongxiao Zhang · University of Fukui Library (University of Fukui) · 2009
This research is concerned with a real-time control system of trajectory tracking and obstacle avoidance using an avoidance manipulability measure for redundant manipulators. To perform predetermined end-effector trajectory-tracking task adaptively without path-planning for avoidance, information on the local environment is naturally restricted by limited recognition time. This means adaptive configuration control has to manage its shape in real-time and without adequate information on its surroundings. Therefore, when the manipulator executes atask adaptively in a dynamic environment, its avoidance manipulability should always be kept as high as possible to prepare for sudden avoidance action. As a measure to gauge the avoidance manipulability based on non-collision, we firstly propose a new index,"AMSIP". By combining a concept of "preview control" with real-time optimization of AMSIP distribution found by "1-step GA", we propose a new real-time configuration control method, with future information being referred locally but effectively. The proposed system has been shown that it is feasible and practical by simulations in terms of real-time configuration optimization. Avoidance manipulability is a new important concept in this research, which is inspired from manipulability. The manipulability represents the ability to generate velocity at the tip of each link without any designated end-effector task. The avoidance manipulability represents the shape-changeability (avoidance ability) of intermediate links when a prior end-effector task is given. Here, the intermediate links denote the all links of the redundant manipulator except the top link with the end-effector since the top link is used to excute the prior task. The avoidance matrix, ^1M_i(i=1,・・・,n-1), is used for analyzing avoidance manipulability of the i-th intermediate link, rank(^1M_i) indicates the shape-changeable space expansion and the singular values of ^1M_i indicate the avoidance ability in the typical direction in the shape-changeable space. As the most essential condition to devise the robot's configuration controller that can always keep the avoidance manipulability high and to build the framework discussing shape-changeability under the prior end-effector task, we analyze what assumption guarantees mathematically the sustainability of the shape-changeable space, that is rank(^1M_i). Then we prove that "Non-Singular Configuration Assumptions" we presented can assure rank(^1M_i) through detailed decomposition analysis of ^1M_i. Non-Singular Configuration Assumptions have not been integrated into our current configuration control system, but they have an ability for presenting yardstick to maintain the sustainability of avoidance space expansion.