Published online by Cambridge University Press: 09 March 2009
A new mathematical formulation of robot and obstacles is presented such that for on-line collision recognition only robot joint positions in workspace are required. This reduces calculation time essentially because joint positions in workspace can be computed every time from the joint variable through robot geometry. It is supposed that the obstacles in the workspace of the manipulator are represented by convex polygons. For every link of the redundant robot and every obstacle a boundary ellipse in 2D is defined in workspace such that there is no collision if the robot joints are outside these ellipses. First, some methods are presented for the automatic determination of these ellipse functions from the obstacle and robot data. Then, the boundary ellipse functions are used as optimization criterion in a collision-avoidance method. The method permits the tip of the hand to follow a given path in the free space while the kinematic control algorithm maximizes the boundary ellipse function of the critical link. The effectiveness of the proposed methods is discussed by theoretical considerations and illustrated by simulations of the motion of three- and four-link manipulators between obstacles.