A New Approach for Collision Avoidance of Manipulators Operating in Unstructured and Time-varying Environments

This paper presents a new approach for collision avoidance of manipulators. When the workspace is completely mapped and the obstacles are known, it is possible to define the task considering the presence of obstacles or their proximity can be monitored simply by the inverse kinematics resolution of the complete system. However, when the workspace is poorly mapped/unstructured and/or time-varying, it is necessary to rely on sensor information to monitor proximity of obstacles and, consequently, define a suitable collision avoidance strategy. This kind of environment is increasingly common in recent robotic applications, and the use of kinematic redundant manipulators is considered, because of increased dexterity. There are being developed different strategies to address this problem, and the literature presents different kinds of sensors and their use to detect the proximity of an obstacle. These strategies use the more traditional pseudo-inverse methods to solve the inverse kinematics of the redundant kinematic chain. In this work, it will be shown the kinematic constraints method as an alternative to solve the inverse kinematics. This method is based on screw theory, the Kirchhoff-Davies method and virtual chains. Among other advantages over the pseudo-inverse methods, the kinematic constraints method is dimensional consistent and conserves movement. An example is used to illustrate the method and its main characteristics.