One of the most important specifications associated with a robot manipulator is repeatability. Unfortunately, there are no standard procedures for determining or specifying the repeatability of a robot. This situation has resulted in a good deal of confusion when attempting to use a published repeatability specification to determine if a robot will meet the requirements of an application.

The purpose of this paper, therefore, is to propose a systematic procedure for the determination and specification of robot repeatability. To accomplish this, three aspects of repeatability analysis are examined. First, a method of collecting data to study repeatability is discussed. A data acquisition system is described as well as an associated test procedure. Next, a procedure for utilizing the acquired data to develop a repeatability specification for a robot is proposed. This procedure results in a specification of the robot's ability to orient objects as well as the classical specification for positioning repeatability. A methodical approach to determining where in the workspace the repeatability test should be made is also addressed. Finally, an approach is presented whereby the data collected during the repeatability test may be used to determine which joints contribute most to the variations in robot repeatability.

The force distribution problem for legged vehicles on rough terrain is considered. A general formulation of the force distribution problem in which the feet contact the ground at arbitrary inclinations, is presented. Three techniques are used to solve the force distribution problem for three representative tasks. The Moore-Penrose pseudo-inverse, a numerical optimization scheme and an approximation to the optimal solution are described. The optimal scheme computes the forces which minimize the maximum ratio of tangential foot reaction force to foot normal force. The approximation is used to achieve certain desirable characteristics of the optimal scheme with considerably less computational resources.

A method of planning crab gaits for a quadruped walking robot is proposed. In the conventional gait study, one of the major concerns has been the foothold selection based on a prescribed body motion, but, in the paper, we consider the body motion planning problem under the assumption that a set of irregular footholds is given, and propose a hierarchical strategy. The strategy can be divided into three stages: first, a feasible range of the body movement is sought under the kinematic limit and the stability constraint. Next, a swing-leg sequence is selected with the aid of a proposed measure of traversability. Finally, an optimal sequence of the body motion is planned by the proposed procedure of otpimizing the gait stability margin. To verify the efficiency of the proposed method, simulation results are presented.

Pattern recognition and object localization, using various sensors such as vision and tactile sensors, are two important areas of research in the application of robotic systems. This paper demonstrates the feasibility of using some relatively inexpensive array of pressure sensors and a neural network approach to achieve object localization and pattern recognition. The sensors that are used are force sensing resistors (FSRs), more specifically, a 16 x 16 array of FSRs. Because of the nonlinearity associated with a FSR, three possible approaches for gathering output from the sensor array are used. The neural network that is used consists of two 2-layer counterpropagation networks (CPNs). One of the CPNs is trained to recognize contact signatures of different objects placed on a fixed reference position on the sensor array.

ISRA is a support system for robot programming. It provides a means to automatically (guided by knowledge) convert a user's request, expressed in the natural language, into the appropriate conceptual model of the required task. This model incorporates information required for the understanding, planning and sensory-guided performance of the required robotic task.

To develop this system we applied the Natural Computation method. We considered natural information processing by humans during synthesis and interpretation of robotic programs, and then constructed an approximate conceptual model of the relevant dynamically changing real world.

Such a model has to be suitable (e.g. representable and executable) for efficient computer processing. This paper presents the results of a case study which shows that methods built by formalizing human information processing in the robotic domain may be efficiently and naturally implemented in ISRA. It provides evidence that the structure and behavior of ISRA's competence representation and algorithm are comparable with the psychological behavior of humans, as required by the Natural Computation method.

The case study illustrates the use of ISRA for the first phase of an assembly program synthesis i.e. planning of all valid assembly sequences.

This paper focuses on developing a planning strategy for robotic assembly. At first, channel and junction are defined by using half-spaces, and free-space inside the female part is approximately decomposed by channels and junctions. Then, a simple and efficient algorithm to find the assembly path of the male part is developed, in which any path between the shortest and safest paths can be easily found by just changing the clearance gap between the male and female parts. Next, the robot arm is considered in the path planning, in which a feasible grasp angle region is obtained to avoid a collision between the robot arm and the female part during the assembly process. An optimum grasping angle can be found in the feasible grasping angle region by applying a proper performance index. Finally, a simple robotic assembly using the algorithm is numerically demonstrated.

The extraction of contour information from objects is essential for purposes of grasping and manipulation. We proposed that human haptic exploration of contours, in the absence of vision, would reveal specialized patterns. Task goals and intrinsic system capacities were assumed to constrain the breadth of processing and the precision with which contour is encoded, thus determining parameters of exploration and ultimately producing movement synergies or “contour exploration procedures.” A methodology for testing these assumptions is described, and the most frequently observed procedures are documented in Part 1. Part 2 will further analyze the procedures, test predictions, and develop implications of the research. The paper (2 parts) is novel in its study of human manipulative behavior from a robotic standpoint; it is thus of interest to robotics research workers interested in the long-term goals of robot manipulation and those interested in an anthropomorphic approach to robotics studies.

Classifications of pictures and pictorial knowledge are presented. Pictorial knowledge is divided into three classes – angular pictorial knowledge, side pictorial knowledge, and angular and side pictorial knowledge. A block diagram of these three pictorial knowledge classes and a pictorial knowledge transformation module is also presented with illustrative examples. Pictorial semantic networks which in terms of pictorial nodes, property nodes, “is a” links, “has property” links, and “if and only if” links are introduced. Transitivity, generalization, specialization, inheritance hierarchy, and knowledge transformation properties are stated and illustrated by examples. Triangular, quadrangular, and polygonal knowledge representation using pictorial semantic networks are presented. The concepts of deducible property nodes are also presented with illustrative examples. Additional facts can be established from pictorial semantic networks. Thus, pictorial semantic networks are a useful way to represent pictorial knowledge in domains that use well-established taxonomies to simplify problem solving in pictorial information systems. Pictorial semantic networks offer what appears to be a fertile field for future study. The results may have useful applications in knowledge representation, expert systems, artificial intelligence, knowledge - based systems, pictorial information systems and related areas.

This paper presents a particular architecture of a parallel robot which is characterized by a diagonal stiffness matrix. This result suggests the use of a parallel robot in the final phase of insertion as a passive compliance device. The stiffness rate of this device is controlled by the gain of the feedback loop. As the correction of small angular misalignments due to contact forces do not generate lateral errors and vice versa, we have the equivalent of an isoelastic swivel.

Present day robot systems are manufactured to perform within industry accepted tolerances. However, to use such systems for tasks requiring high precision, various methods of robot calibration are generally required. These procedures can improve the accuracy of a robot within a small volume of the robot's workspace. The objective of this paper is to demonstrate the use of a single camera 3D computer vision system as a position sensor in order to perform robot calibration. A vision feedback scheme, termed Vision-guided Robot Control (VRC), is described which can improve the accuracy of a robot in an on-line iterative manner. This system demonstrates the advantage that can be achieved by a Cartesian space robot control scheme when end effector position/orientation are actually sensed instead ofcalculated from the kinematic equations. The degree of accuracy is determined by setting a tolerance level for each of the six robot Cartesian space coordinates. In general, a small tolerance level requires a large number of iterations in order to position the end effector, and a large tolerance level requires fewer iterations. The viability of using a vision system for robot calibration is demonstrated by experimentally showing that the accuracy of a robot can be drastically improved. In addition, the vision system can also be used to determine the repeatability and accuracy of a robot in a simple, efficient, and quick manner. Experimental work with an IBM Electric Drive Robot (EDR) and the proposed vision system produced a 97 and a 145 fold improvement in the position and orientation accuracy of the robot, respectively.