Obstacle detection is a critical task in mobile robot navigation. To allow for rapid reactions to changes in the environment, it is required to complete its operation in short time intervals. That can be difficult if computing and/or sensing robot resources are limited. Therefore, it is important to design precisely what sensorial information would be needed to guarantee detection, and how it would be gathered to guarantee throughput. In this paper we present analysis and experiments related to the design of an obstacle detection system for a mobile robot with limited sensing capabilities. Objectives are to preserve safety while maintaining good motion performance. The design is made in two steps. Firstly, to find what is the minimum information needed to assure safety as a function of the sensors at hand, the robot capabilities and the intended motion control policy. Second, to manage the available sensors in order to provide the required information when needed. For the example studied, a justified solution has been defined which guarantees detection, robot safety, and good motion performance in real time. The solution has been tested with experiments that illustrate the effectivity achieved.

This paper describes a method for analyzing undulation and off-center of a flat product in a web inspection by a linear camera. The longitudinal undulation in the plate appears to the high speed of the line and different tensions across the strip section as it is pulled to the end of the line. If the amplitude and/or frequency of the undulation is too high it will produce an increase of breakage probability and new defects may appear. In addition, it introduces errors in width measurement if it is not taken into account. We propose a method to identify undulation characteristics based on time-frequency analysis of the projection of the strip edge. We recover the position of the edge and undulation by assuming that band off-center is a slower process than undulation. Simulations with different undulation waveforms and off-centers, as well as experimental results are shown.

Robot Navigation is an especially challenging problem when only online sensor information is available. The main problem is to guarantee global properties, such as algorithm convergence or trajectory optimality, based on local information. In this paper we present a new non-heuristic sensor-based planning algorithm, characterized by: 1) it is based in potential functions, allowing to introduce optimality criteria, 2) it is computed incrementally to introduce last sensor readings, and 3) it accounts for robot dynamics. The result is a method suitable for real-time navigation, it is intuitive and easy to understand, and produces smooth and safe trajectories.

In experimental robotics it is common to complement the motion planning algorithm with a local obstacle avoidance module. We are concerned here with the specifications of a range sensor specifically design for the task. Such sensor has special requirements: it has to guarantee detection and response, and it has to guarantee throughput, providing information ail a rate in proportion to robot velocity. An analysis is presented which allows the selection of the sensor requirements for collision avoidance tasks in mobile robots. The design takes into account robot dynamics, it is compatible with fast motion-only the indispensable environment zones are explored-and avoids unnecessary velocity reductions. Experiments show the possibilities and limitations of the approach.

The feasibility of realistic autonomous space manipulation tasks using multisensory information is shown through two experiments involving a fluid interchange system and a module interchange system. In both cases, autonomous location of the mating element, autonomous location of the guiding light target, mating, and demating of the system are performed. Specifically, vision-driven techniques were implemented that determine the arbitrary two-dimensional position and orientation of the mating elements as well as the arbitrary three-dimensional position and orientation of the light targets. The robotic system is also equipped with a force/torque sensor that continuously monitors the six components of force and torque exerted on the end effector. Using vision, force, torque, proximity, and touch sensors, the fluid interchange system and the module interchange system experiments were accomplished autonomously and successfully.>

A general approach is presented for the integration of vision, range, proximity, and touch sensory data to derive a better estimate of the position and orientation (pose) of an object appearing in the work space. Efficient and robust methods for analyzing vision and range data to derive an interpretation of input images are discussed. Vision information analysis includes a model-based object recognition module and an image-to-world coordinate transformation module to identify the three-dimensional (3-D) coordinates of the recognized objects. The range information processing includes modules for reprocessing, segmentation, and 3-D primitive extraction. The multisensory information integration approach represents sensory information in a sensor-independent form and formulates an optimization problem to find a minimum-error solution to the problem. The capabilities of a multisensor robotic system are demonstrated by performing a number of experiments using an industrial robot equipped with several sensors of differing types.>

The authors describe research efforts directed towards the development of sensor-driven robotic systems for performing various inspection and manipulation tasks. Emphasis is on modules designed for analyzing vision and range sensor data for object recognition and pose estimation. The object recognition system uses a model-based approach. Pose estimation by vision sensor includes derivation of an image-to-world coordinate transformation module. Range data analysis includes modules for preprocessing, segmentation, and 3-D primitive extraction. Experiments are presented to demonstrate the applicability of the analysis modules using a laboratory-based robotic testbed.>

The location of a scene element can be determined from the disparity of two of its depicted entities (each in a different image). Prior to establishing disparity, however, the correspondence problem must be solved. It is shown that for the axial-motion stereo camera model the probability of determining unambiguous correspondence assignments is significantly greater than that for other stereo camera models. However, the mere geometry of the stereo camera system does not provide sufficient information for uniquely identifying correct correspondences. Therefore, additional constraints derived from justifiable assumptions about the scene domain and from the scene radiance model are utilized to reduce the number of potential matches. The measure for establishing the correct correspondence is shown to be a function of the geometrical constraints, scene constraints, and scene radiance model.>

A thresholding technique is developed for segmenting digital images with bimodal reflectance distributions under nonuniform illumination. The algorithm works in a raster format, thus making it an attractive segmentation tool in situations requiring fast data throughput. The theoretical base of the algorithm is a recursive Taylor expansion of a continuously varying threshold tracking function.

A new closed-form expression for the semi-invariants of the interclass Mahalanobis distance is derived. Typically, in the analysis of two multivariate Gaussian populations with different covariance matrices, simultaneous diagonalization of these matrices is required. The semi-invariants are given directly in terms of the mean vectors and inverse covariance matrices by the results established. In addition, a new iterative algorithm is derived for computing the moments of the interclass Mahalanobis distance from the semi-invariants.

The worst-case number of variable evaluations (testing cost) of Boolean functions is examined. Following up on a result by Rivest and Vuillemin, we show that all symmetric as well as all linearly separable Boolean functions are exhaustive, that is, have a pessimal worst-case testing cost.

It is shown that the moments of the interclass Mahalanobis distance between elements of twod-variate Gaussian populations can be expressed in a simple polynomial form. Thenth moment is expressible as a polynomial of ordernwhose variable depends on the mean vectors and eigenvalues of the covariance matrices. A closed-form solution is given for computing the coefficients of the polynomial expressions.

This paper examines the amplitude fluctuations of band-limited functions with bounded zeroth absolute moments, and the problems associated with estimating the level crossing profiles of these functions. Level crossings have received increased attention as features for pattern recognition because of their capability to provide information related to both amplitude and frequency behavior. A detailed analysis of the average rate of change of band-limited functions is presented, including a derivation of the least upper bound on functions for which the zeroth absolute moments of the functions are bounded. The average rate of change of a function over an interval in which one endpoint of the interval is an extremum of the function is similarly bounded and used to establish a sampling rate which guarantees that between successive samples of a band-limited function with bounded zeroth absolute moment the function itself does not deviate from the amplitude interval defined by the samples by more than some predefined amplitude change. Based on these results, a theorem is developed which defines the sampling rate required to ensure that, in the estimation of level crossing profiles, no more than 2m (m >> 1) level crossings of m levels are missed per extremum of the sampled function. It is shown that the sampling rate defined by this theorem reduces to the well-known Nyquist rate for the special case of zero crossing analysis.

The equivalence of nonterminals of an expansive tree grammar is considered. Algorithms are presented for constructing sets of equivalent nonterminals for an expansive tree grammar, for reducing the grammar, and for determining whether two grammars generate the same language.

Pattern recognition techniques were applied to the computation of void fraction values in two-phase steam/water mixtures. The void fraction is one of the principal parameters used to characterize the heat transfer properties of two-phase flows in the emergency cooling system of pressurized water reactors. The recognition approach reported here is based on moment descriptors extracted from the amplitude signal of an impedance probe. Using data obtained at the Advanced Instrumentation for Reflood Studies facility of the Oak Ridge National Laboratory, it was shown that void fraction values could be consistently recognized with less than 10 percent error using data segments as short as 0.25 s in duration.

The construction of sequential testing procedures from functions of discrete arguments is a common problem in switching theory, software engineering, pattern recognition, and management. The concept of the activity of an argument is introduced, and a theorem is proved which relates it to the expected testing cost of the most general type of decision trees. This result is then extended to trees constructed from relations on finite sets and to decision procedures with cycles. These results are used, in turn, as the basis for a fast heuristic selection rule for constructing testing procedures. Finally, some bounds on the performance of the selection rule are developed.

The algorithm presented in this paper is a direct approach to the problem of determining a linear discriminant function which minimizes total classification error in a two-class training set. Discriminant function coefficients are determined by a search technique which is based on two theorems derived from convexity theory. The number of iterative steps in the search sequence has a theoretical upper bound which is strictly less than the number required for full enumeration, and applications of the algorithm to experimental data have established its tractability under nontrivial conditions. Although the training set patterns are initially assumed to be in general position, this restriction is essentially eliminated by an extension of the algorithm which is briefly discussed.

A multivariate statistical pattern recognition system for reactor noise analysis is presented. The basis of the system is a transformation for decoupling correlated variables and algorithms for inferring probability density functions. The system is adaptable to a variety of statistical properties of the data, and it has learning, tracking, updating, and dimensionality reduction capabilities. System design emphasizes control of the false-alarm rate. Its abilities to learn normal patterns and to recognize deviations from these patterns were evaluated by experiments at the Oak Ridge National Laboratory (ORNL) High-Flux Isotope Reactor. Power perturbations of less than 0.1 percent of the mean value in selected frequency ranges were readily detected by the pattern recognition system.

The methods developed in this correspondence represent an approach to the problem of handling error-corrupted syntactic pattern strings, an area generally neglected in the numerous techniques for linguistic pattern description and recognition which have been reported. The basic approach consists of applying error transformations to the productions of context-free grammars in order to generate new grammars (also context-free) capable of describing not only the original error-free patterns, but also patterns containing specific types of errors such as deleted, added, and interchanged symbols which arise often in the pattern-scanning process. Theoretical developments are illustrated in the framework of a syntactic recognition system for chromosome structures.

An algorithm for the optimal solution of consistent and inconsistent linear inequalities is presented, where the optimality criterion is the maximization of the number of satisfied constraints. The algorithm is developed as a nonenumerative search procedure based on two new theorems established in this paper. It is shown that the number of iterative steps before termination is strictly less than that required by an exhaustive search. Experimental results with various types of data establish the computational tractability of the procedure under nontrivial conditions.

A handwritten character recognition system has been designed by making use of topological feature extraction and multilevel decision making. By properly specifying a set of easily detectable topological features, it is possible to convert automatically the handwritten characters into stylized forms and to classify them into primary classes with similar topological configurations. Final recognition is accomplished by a secondary stage that performs local analysis on the characters in each primary category. The recognition system consists of two stages: global recognition, followed by local recognition. Automatic character stylization results in pattern clustering which simplifies the classification tasks considerably, while allowing a high degree of generality in the acceptable writing format. Simulation of this scheme on a digital computer has shown only 6 percent misrecognition.