Manipulation Task Planning and Motion Control Using Task Relaxations

When describing high-level manipulation task specifications, it is common to discretize the workspace in multiple locations to define different possibilities for placing the manipulated objects, which usually does not scale well. Hence, we use a kinematic control law based on constrained quadratic programming to relax tasks by defining regions of interest instead of exact locations, with geometric primitives represented by dual quaternions, in conjunction with a multi-layered framework for task and motion planning already available in the literature. The framework consists of a high-level planner, which generates task plans for linear temporal logic (LTL) specifications, and of a low-level motion controller, which executes the task.