Critical configuration path planning and knowledge-based task planning for robot systems
Abstract
This research investigates the path planning problem and the task planning problem for robot systems. A critical configuration algorithm has been developed to solve the rotation problem, which is crucial to the path planning problem. By considering the rotations of the moving object, a set of critical configurations can be defined. Then, according to the geometric properties of the environment, a set of trigonometric equations can be obtained. Solving these equations can acquire the value of $a\sp\*$, which is defined as the maximal length that is allowed for the object to pass a corner when the other parameters are fixed. Consequently, comparing the value of $a\sp\*$ to the longitude of the moving object, the existence of a safe path can be decided. Based on the principle of divide-and-conquer, a 3-D motion planning algorithm for a Cartesian manipulator has also been developed in this thesis. A collision-free path can be found in a systematic way by dividing the manipulator motion into three submotions: a departure motion, an intermediate motion, and an approach motion. With a distinct 3-D grid representation for a workspace, the problem can be transformed into a 2-D problem. First, a temporary solution is obtained. Subsequently, according to some heuristic information of the environment, a near optimal solution can be acquired. By using the geometric information of an environment, we further formulate the robot task planning problem as a find-path problem. The basic idea is that a plan exists if and only if the robot can find collision-free paths to accomplish the goal. Due to the recent advances of object-oriented systems and knowledge-based systems, we are able to transform a high-level goal statement into a set of object-oriented tasks and then a set of robot-level operations. Consequently, by incorporating proper path planning algorithms, the robot task planning problem can be solved in polynomial time.
Degree
Ph.D.
Advisors
Lee, Purdue University.
Subject Area
Electrical engineering|Artificial intelligence
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.