James S. Tittle

Envisioning human-robot coordination in future operations

Developers of autonomous capabilities underestimate the need for coordination with human team members when their automata are deployed into complex operational settings. Automata are brittle as literal minded agents and there is a basic asymmetry in coordinative competencies between people and automata. The new capabilities of robotic systems raise new questions about how to support coordination. This paper presents a series of issues that demand innovation to achieve human-robot coordination (HRC). These include supporting people in their roles as problem holder and as robotic handler, overcoming ambiguities in remote perception, avoiding coordination surprises by better tools to see into future robotic activities and contingencies, and responsibility in human-robot teams.

DISTORTIONS OF THREE-DIMENSIONAL SPACE IN THE PERCEPTUAL ANALYSIS OF MOTION AND STEREO

A fundamental issue in the study of human vision is the accuracy with which observers can perceive the three-dimensional structures of objects in the environment. The formal mapping from physical to perceived space is considered as a geometric transformation, and the literature is reviewed to identify which aspects of object structure are systematically distorted by this transformation and which ones remain invariant. In the perceptual analysis of several different sources of optical information, including motion and stereo, both individually and in combination, there is a consistent pattern of results to indicate that perceived depth intervals are scaled differently from comparable intervals in either horizontal or vertical directions. These and other findings provide strong evidence that the relationship between physical and perceived space is noneuclidean, and that the three-dimensional structures of objects can appear systematically distorted even when viewed under full cue conditions.

Recovery of 3-D shape from binocular disparity and structure from motion

Four experiments were conducted to examine the integration of depth information from binocular stereopsis and structure from motion (SFM), using stereograms simulating transparent cylindrical objects. We found that the judged depth increased when either rotational or translational motion was added to a display, but the increase was greater for rotating (SFM) displays. Judged depth decreased as texture element density increased for static and translating stereo displays, but it stayed relatively constant for rotating displays. This result indicates that SFM may facilitate stereo processing by helping to resolve the stereo correspondence problem. Overall, the results from these experiments provide evidence for a cooperative relationship between. SFM and binocular disparity in the recovery of 3-D relationships from 2-D images. These findings indicate that the processing of depth information from SFM and binocular disparity is not strictly modular, and thus theories of combining visual information that assume strong modularity-or-independence cannot accurately characterize all instances of depth perception from multiple sources.

Recovering viewer-centered depth from disparity, occlusion, and velocity gradients

Two experiments were conducted to assess the effects of corresponding and conflicting binocular and monocular information on the recovery of depth order (signed depth). Subjects viewed displays in which the same or opposite depth orders were indicated by disparity and occlusion, in one experiment, or by disparity and velocity gradients, in a second experiment. The same 36 subjects, 17 who had failed a Random Dot E test and 19 who had passed, were run in both experiments. When binocular and monocular information indicated conflicting depth orders, most subjects responded in accordance with the monocular information on some trials in both experiments. This was true even for a subgroup who always responded in accordance with the stereoscopic information on control trials that did not provide monocular information for depth order. For this subgroup, the impact of conflicting monocular information in the velocity gradient task correlated with performance on the uncrossed version of the Random Dot E test. We also found that some subjects who failed static tests of stereoscopic depth perception could respond accurately to continuously changing disparities.

The observer-relative velocity field as the basis for effective motion parallax.

Earlier studies of motion parallax found unambiguous relative depth perception when random dot patterns were systematically translated in accordance with either motion of the observers head or motion of the display scope. The need for such relative motion between an observer and a flow field was examined by placing a flow field in a limited area (window) in a large scope and translating the window relative to the observer. Accuracy in judging surface orientation and quantitative depth estimates were determined by the velocity field relative to the observer and were not measurably affected by whether this field was produced with a stationary or a moving window. Accuracy was consistently higher for smaller ratios of maximum to minimum projected velocities, reaching 100% in one experiment with a 1.12:1 ratio. We conclude that fully effective motion parallax does not require relative motion between the observers head and the contours of a flow field.

The Remote Perception Problem

Previous research (e.g., Casper, 2002; Darken, Kempster, & Peterson, 2001) has shown that observers demonstrate poor spatial awareness based on video provided from remote environments. Such a result is understandable given that remote vision systems provide impoverished representations that leave out higher order cues essential to build coherent percepts and models of the world being explored. If tele-presence or remote vision is to be useful in the future, the raw video needs to somehow be augmented to recover what was lost by decoupling the human perceptual processor from the natural environment.

FINDING DECISION SUPPORT REQUIREMENTS FOR EFFECTIVE INTELLIGENCE ANALYSIS TOOLS

Within ARDAs GI2Vis program, we developed a unique framework for the definition of decision support requirements for intelligence analysis tools. This framework, based on a first-of-a-kind integration of a model of inferential analysis and principles for designing effective human-computer teams from Cognitive Systems Engineering, has defined the essential support functions to be provided to the intelligence analyst(s). This model has proven to be extremely useful in assessing the support provided by a large set of visualization tools. This assessment has identified clusters of support functions that are addressed by many tools as well as key missing support functions. In this way, the Support Function Model has been used to identify gaps in the support function coverage of existing tools. This can serve as a valuable focusing mechanism for future design and development efforts. In addition, we believe this would be a useful mechanism to enhance cross-discussions among research teams involved in Cognitive Task Analysis efforts within the Intelligence Community. Having others integrate their analytic results with this framework would provide the mechanism for expansion of this model to become a more robust tool and have an even greater impact on the Intelligence Community.

Recovering three-dimensional shape from perspective translations and orthographic rotations

Perceived orientation in depth and 3D shape was investigated for perspective projections of translations and orthographic projections of rotations of 3D dihedral angles. The principal findings were that (a) perceived orientation in depth depends on the sign of the velocity gradient, even in the case of orthographic projections; (b) the relationship between perceived orientation and the sign of the velocity gradient is greater for shallower gradients in orthographic projections of rotations, consistent with previous findings for perspective translations; (c) the magnitudes of simulated dihedral angles were underestimated (relative depth overestimated) for orthographic projections of rotations but were overestimated for perspective projections of translations; and (d) the judged magnitude of the dihedral angle depends on the velocity ratio and on image compression; it cannot be predicted from the velocity ratio or the velocity gradient alone.

The Perception of Scale-Dependent and Scale-Independent Surface Structure from Binocular Disparity, Texture, and Shading

The integration of binocular disparity, shading, and texture was measured for two different aspects of three-dimensional structure: (1) shape index, which is a measure of scale-independent structure, and (2) curvedness, which is a measure of scale-dependent structure. Binocular disparity was found to contribute significantly more to judged shape index than it does to judged curvedness, and shading and texture were both found to contribute more to judged curvedness than to judged shape index. These results demonstrate that different cues do not contribute equally to different aspects of perceived surface structure. This finding suggests that, for the case of linear integration, multiple cues to three-dimensional structure do not combine on the basis of a single type of representation shared by all the ‘shape-from-X’ processes in the visual system.

Stereoscopic depth perception by static stereo-deficient observers in dynamic displays with constant and changing disparity.

The performance of 11 static stereo-deficient subjects and 11 static stereo-normal subjects was compared on two types of dynamic stereo displays-one where disparities were constant during motion and one where disparities changed continuously. Computer-generated displays simulating horizontal motion of figures at different depths or rotation of figures about a vertical axis were viewed through a Brewster stereoscope. About one-half of the subjects in our static stereo-deficient sample were able to make depth judgments on the basis of disparity in both types of dynamic displays. The clinical feature which appeared to distinguish those static stereo-deficient subjects who could use disparity information in dynamic displays from those who could not was early onset constant strabismus. These results indicate that a complete evaluation of stereo ability should include tests with dynamic displays, possibly including both constant and changing disparities.