Note: Please contact me or email@example.com if you need a paper not accessible on-line.
Abstract : Automatic recognition of unknown contacts among objects provides important high-level feedback for a reactive robot system, especially a robot system for manipulating or assembling objects with high-precision requirements. The recognition mechanism must be on-line and insensitive to sensing uncertainties. To achieve this, different sensors compensating one another should be used, such as position/orientation and vision sensors. In this paper, we introduce a method of on-line using vision to extract more accurate contact information by reducing a set of initial contact hypotheses which are obtained from position/orientation sensing and subject to sensing uncertainties. Currently we assume contacts are among polyhedral objects with known geometric models. Preliminary implementation of the algorithms are discussed with testing results shown.
Abstract : This paper provides a general and exact method of growing a polyhedral object in three-dimensional Cartesian space to take into account its orientation and position uncertainties. The work is particularly motivated by the need for automatically recognizing contact situations among objects in spite of uncertainties. The technique of growing surface elements of a polyhedral by uncertainty can be used to extract the set of all possible contact situations among polyhedral objects in the presence of location uncertainties, which can then serve as a basis for further and more accurate extraction of contact information by additional sensing means, such as vision and force/moment sensing.
Abstract : Sensing, modeling, control, and manufacturing uncertainties make it generally impossible to guarantee the success of high-precision robot tasks. This paper analyzes the relationships among these uncertainty parameters (including a robot, parts, and sensors) and derives design and motion constraints whose satisfaction can allow a robot motion, and consequently, a task to succeed. By using a cylindrical peg-in-hole task as an example, quantitative conditions are obtained under which the use of force/moment sensing allows position sensing tolerances to be significantly relaxed. Simulation results uphold the theoretical derivations and show empirically that the theoretical constraints can be relaxed somewhat with good results still obtained.
Abstract : A key problem in robotics application on high-precision tasks, such as assembly tasks, is how to make robots operate reliably in the presence of uncertainties (such as mechanical, control, and sensor uncertainties). Since there is no general and unconditional solution for the problem, the uncertainty handling for robot assembly must be a dynamic process involving sensory information and general knowledge of contacts among the parts being assembled, and its success can only be guaranteed if certain constraints on the nominal design parameters, tolerances, and sensor error parameters are enforced. Based on the above belief, this dissertation introduces a replanning approach towards uncertainty handling by presenting a task-independent replanning strategy based on knowledge of contact and sensory data, and showing how eventual success of a task can be guaranteed in spite of certain class of sensor, control and manufacturing imperfections if certain design and motion constraints are satisfied.
Abstract : High-precision assembly tasks cannot be successfully done by robots without taking into account the effect of uncertainties. Often a robot motion may fail and result in some unintended contact between the object held by the robot and the environment. Aiming at automatically recovering a task from such a failure, we earlier introduced a replanning approach consisting of patch-planning based on knowledge of contact and motion strategy planning . This paper presents contact-based general strategies to create rotational patch-plans for correcting orientational errors of the held object automatically.
Abstract : In robotic operations, it is often desired to use compliant motions to keep a moving object (held by a manipulator hand) in contact with, or constrained by its environment. There have been considerable studies on implementing compliant motions. However, little attention has been given to the issue of ensuring successful implementation in the presence of uncertainties. This paper studies the relationships among the commanded force/moment and the sensing, modeling and control uncertainties, and shows that when certain constraints are satisfied, compliant translations/rotations can be ensured in spite of uncertainties.
Abstract : An Evolutionary Algorithm is discussed for the path planning problem in mobile robot environment, which may contain a number of unknown obstacles. The Evolutionary Algorithm searches for paths in the entire, continuous free space. It unifies off-line and on-line planning processes and provides high safety measures without requiring complete information about the obstacles sensed.
Abstract : In any evolutionary system, the evaluation function for rating the fitness of solutions serves as the major link between the problem and the algorithm. A definition of the evaluation function directly affects the effectiveness and the efficiency of the algorithm. In this paper we study this issue in the context of the Evolutionary Planner/Navigator (EP/N), an evolutionary system which we have developed for mobile robot planning and navigation. The novel design of the chromosome structure and the genetic operators (customized to this special application) makes the EP/N efficient, flexible to changes and robust to uncertainties. On the basis of such design, it is the evaluation function of the EP/N which accommodates different optimization needs and indirectly affects the course of the planning process. Hence, much of our research effort on the EP/N is on studying how different evaluation functions affect the quality of results. The insights gained from this research may be applicable to other evolutionary systems as well.