Vehicle manufacturers and government agencies are considering using vehicle-to-pedestrian (V2P) communication to improve pedestrian safety. However, there are unanswered questions about whether people will heed alerts and warnings presented through a smartphone. We conducted between-subject studies with younger and older adults where they physically crossed a virtual street. They received either permissive alerts (safe to cross), prohibitive warnings (not safe to cross), or no alerts or warnings (control). We found that both older and younger adults were highly likely to heed permissive alerts, even when this meant taking gaps between two vehicles that were smaller than they would typically take on their own. We also found that we could shift participants’ road-crossing behavior toward greater caution when they were only alerted to cross very large gaps between two vehicles. Participants stated that alerts and warnings were useful, but that prohibitive warnings were annoying. These findings give insights into V2P design and pedestrian behavior when smartphone assistance is provided.

This study investigated how people coordinate their decisions and actions with a risky or safe computer-generated agent in a humanoid or non-humanoid form and how this experience influences later behavior when acting alone. In Experiment 1, participants first repeatedly crossed continuous traffic in a virtual environment with a humanoid computer-generated agent (Figure 1). Participants were specifically instructed to cross with an agent that was programmed to be either safe (taking only large gaps) or risky (also taking relatively small gaps). Participants then repeatedly crossed the same roadway alone. We found that participants’ experiences with crossing safe vs. risky gaps with an agent persisted in later trials when the participants crossed alone, such that participants accepted tighter gaps if they were previously paired with a risky than a safe agent. In Experiment 2 (Figure 2), we tested whether experience crossing with a risky or safe non-humanoid object (a floating box) also influenced later behavior when crossing alone. We again found that participants who crossed with the risky object partner took tighter gaps when later crossing alone than those who crossed with the safe object partner. The Discussion focuses on the impact of experiences with virtual agents on perception–action tuning and the potential of using virtual agents for training safe road-crossing behavior.

This paper examines how people jointly coordinate their decisions and actions with risky vs. safe human and agent road-crossing partners (Fig. 1 ). The task for participants was to physically cross a steady stream of traffic in a large-screen virtual environment without getting hit by a car. The computer-generated (CG) agent was programmed to be either safe (taking only large gaps) or risky (also taking small gaps). The human partners were classified as safe (taking more large gaps) or risky (also taking some small gaps) based on their average gap size selection. We found that participants in all four conditions preferred to cross with their partner. As a consequence, the riskiness of the partner (both human and agent) influenced the riskiness of participants' gap choices. We also found that participants tightly synchronized their movement with both human and agent partners. The largest differences in performance between those paired with agent vs. human partners occurred on trials when participants crossed different gaps than their partners. This study demonstrates the potential for studying how people interact with CG agents when performing whole-body joint actions using large-screen immersive virtual environments.

We investigated how two people jointly coordinate their decisions and actions in a co-occupied, large-screen virtual environment. The task for participants was to physically cross a virtual road with continuous traffic without getting hit by a car. Participants performed this task either alone or with another person (see Figure 1). Two separate streams of non-stereo images were generated based on the dynamic locations of the two viewers’ eye-points. Stereo shutter glasses were programmed to display a single image stream to each viewer so that they saw perspectively correct non-stereo images for their eyepoint. We found that participant pairs often crossed the same gap together and closely synchronized their movements when crossing. Pairs also chose larger gaps than individuals, presumably to accommodate the extra time needed to cross through gaps together. These results demonstrate how two people interact and coordinate their behaviors in performing whole-body, joint motions in a co-occupied virtual environment. This study also provides a foundation for future studies examining joint actions in shared VEs where participants are represented by graphic avatars.

This paper investigates the influence of stereoscopic vs. non-stereoscopic display in large-screen virtual environments on an everyday perception-action task - crossing traffic-filled roadways as a pedestrian. The task for participants was to physically cross a virtual road with continuous traffic without getting hit by a car in a CAVE-like virtual environment. Half of the participants performed the task with stereoscopic display and half performed the task with non-stereoscopic display. We found that stereoscopic display had little impact on the size of the gaps participants crossed or the timing of their crossing motion relative to the gap with the exception of a small difference in crossing speed. The results are important for validating the use of non-stereoscopic image displays in ground vehicle simulation and supporting the use of non-stereoscopic displays for multi-viewpoint rendering in co-occupied virtual environments.

This investigation compared how people performed a complex perception-action task - crossing traffic-filled roadways - in a CAVE vs. an HMD virtual environment. Participants physically crossed a virtual roadway with continuous cross traffic in either a CAVE-like or an HTC Vive pedestrian simulator. The 3D model and traffic scenario were identical in both simulators, allowing for a direct comparison between the two display systems. We found that participants in the Vive group accepted smaller gaps for crossing than participants in the CAVE group. They also timed their entry into the gap more precisely and tended to cross somewhat more quickly. As a result, participants in the Vive group had a somewhat larger margin of safety when they exited the roadway than those in the CAVE group. The results provide a foundation for future studies of pedestrian behavior and other tasks involving full-body motion using HMD-based VR.

Optical motion capture is based on estimating the three-dimensional positions of markers by triangulation from multiple cameras. Successful performance depends on points being visible from at least two cameras and on the accuracy of the triangulation. Triangulation accuracy is strongly related to the positions and orientations of the cameras. Thus, the configuration of the camera network has a critical impact on performance. A poor camera configuration may result in a low quality three-dimensional (3D) estimation and consequently low quality of tracking. This paper introduces and compares two methods for camera placement. The first method is based on a metric that computes target point visibility in the presence of dynamic occlusion from cameras with "good" views. The second method is based on the distribution of views of target points. Efficient algorithms, based on simulated annealing, are introduced for estimating the optimal configuration of cameras for the two metrics and a given distribution of target points. The accuracy and robustness of the algorithms are evaluated through both simulation and empirical measurement. Implementations of the two methods are available for download as tools for the community.

We investigated how people jointly coordinate their decisions and actions with a computer-generated character (agent) in a large-screen virtual environment. The task for participants was to physically cross a steady stream of traffic on a virtual road without getting hit by a car. Participants performed this task with another person or with a computer-generated character (Fig. 1). The character was programmed to be either safe (taking only large gaps) or risky (also taking relatively small gaps). We found that participants behaved in many respects similarly with real and virtual partners. They maintained similar distances between themselves and their partner, they often crossed the same gap with their partner, and they synchronized their crossing with their partner. We also found that the riskiness of the character influenced the gap choices of participants. This study demonstrates the potential for using large-screen virtual environments to study how people interact with CG characters when performing whole-body joint actions.

We investigated how two people jointly coordinate their decisions and actions in a co-occupied, large-screen virtual environment. The task for participants was to physically cross a virtual road with continuous traffic without getting hit by a car. Participants performed this task either alone or with another person (see Fig.1). We found that pairs often crossed the same gap together and closely synchronized their movements when crossing. Pairs also chose larger gaps than individuals to accommodate the extra time needed to cross through gaps together. These results reveal how two people interact and coordinate their behaviors in performing whole-body, joint motions. This study also provides a foundation for future studies examining joint actions in shared VEs where participants are represented by graphic avatars.

This paper presents an experiment conducted in a large-screen immersive virtual environment to evaluate how texting pedestrians respond to permissive traffic alerts delivered via their cell phone. We developed a cell phone app that delivered information to texting pedestrians about when traffic conditions permit safe crossing. We compared gap selection and movement timing in three groups of pedestrians: texting, texting with alerts, and no texting (control). Participants in the control and alert groups chose larger gaps and were more discriminating in their gap choices than participants in the texting group. Both the control and alert groups had more time to spare than the texting group when they exited the roadway even though the alert group timed their entry relative to the lead car less tightly than the control and texting groups. By choosing larger gaps, participants in the alert group were able to compensate for their poorer timing of entry, resulting in a margin of safety that did not differ from those who were not texting. However, they also relied heavily on the alert system and paid less attention to the roadway. The discussion focuses on the potential of assistive technologies based on Vehicle-to-Pedestrian (V2P) communications technology for mitigating pedestrian-motor vehicle crashes.

Optical motion capture is based on estimating the three-dimensional positions of markers by triangulation from multiple cameras. Successful performance depends on points being visible from at least two cameras and on the accuracy of the triangulation. Triangulation accuracy is strongly related to the positions and orientations of the cameras. Thus, the configuration of the camera network has a critical impact on performance. A poor camera configuration may result in a low quality three-dimensional (3D) estimation and consequently low quality of tracking. This paper proposes a camera configuration metric that is based on a probabilistic model of target point visibility from cameras with "good" views. An efficient algorithm, based on simulated annealing, is introduced for estimating the optimal configuration of an arbitrary set of cameras for a given distribution of target points. The accuracy and robustness of the algorithm are evaluated through both simulation and empirical measurement. An implementation of the method is available for download as a tool for the community [Rahimian and Kearney 2015] (Figure 1).

We investigated how the properties of interactive virtual reality systems affect user behavior in full-body embodied interactions. Our experiment compared four interactive virtual reality systems using different display types (CAVE vs. HMD) and modes of locomotion (walking vs. joystick). Participants performed a perceptual-motor coordination task, in which they had to choose among a series of opportunities to pass through a gate that cycled open and closed and then board a moving train. Mode of locomotion, but not type of display, affected how participants chose opportunities for action. Both mode of locomotion and display affected performance when participants acted on their choices. We conclude that technological properties of virtual reality system (both display and mode of locomotion) significantly affected opportunities for action available in the environment (affordances) and discuss implications for design and practical applications of immersive interactive systems.

We conducted a series of experiments to investigate effects of scale changes on distance perception in virtual environments. All experiments were carried out in an HMD. Participants first made distance estimates with feedback in a virtual tunnel (adaptation) and then made distance estimates without feedback in a differently-scaled virtual environment (test). We examined several types of scale changes, including changing the size of (1) the tunnel, (2) the targets, and (3) the separation of the two targets. Changes in target size always affected distance estimates at test. When the targets became smaller, participants overshot distance and when the targets became larger, participants undershot distance. Changes in the size of the tunnel or the separation between the targets (without a change in the size of the targets) had a minimal effect on distance estimates. These results indicate that distance estimates at test were strongly influenced by familiar size cues for distance. The discussion focuses on the stability of calibration processes and mechanisms for cue integration for perceiving distance in virtual environments.

The goal of our work is to develop a programmatically controlled peer to bicycle with a human subject for the purpose of studying how social interactions influence road-crossing behavior. The peer is controlled through a combination of reactive controllers that determine the gross motion of the virtual bicycle, action-based controllers that animate the virtual bicyclist and generate verbal behaviors, and a keyboard interface that allows an experimenter to initiate the virtual bicyclist's actions during the course of an experiment. The virtual bicyclist's repertoire of behaviors includes road following, riding alongside the human rider, stopping at intersections, and crossing intersections through specified gaps in traffic. The virtual cyclist engages the human subject through gaze, gesture, and verbal interactions. We describe the structure of the behavior code and report the results of a study examining how 10- and 12-year-old children interact with a peer cyclist that makes either risky or safe choices in selecting gaps in traffic. Results of our study revealed that children who rode with a risky peer were more likely to cross intermediate-sized gaps than children who rode with a safe peer. In addition, children were significantly less likely to stop at the last six intersections after the experience of riding with the risky than the safe peer during the first six intersections. The results of the study and children's reactions to the virtual peer indicate that our virtual peer framework is a promising platform for future behavioral studies of peer influences on children's bicycle riding behavior.

We conducted two experiments that compared distance perception in real and virtual environments in six visual presentation methods using either timed imagined walking or direct blindfolded walking, while controlling for several other factors that could potentially impact distance perception. Our presentation conditions included unencumbered real world, real world seen through an HMD, virtual world seen through an HMD, augmented reality seen through an HMD, virtual world seen on multiple, large immersive screens, and photo-based presentation of the real world seen on multiple, large immersive screens. We found that there was a similar degree of underestimation of distance in the HMD and large-screen presentations of virtual environments. We also found that while wearing the HMD can cause some degree of distance underestimation, this effect depends on the measurement protocol used. Finally, we found that photo-based presentation did not help to improve distance perception in a large-screen immersive display system. The discussion focuses on points of similarity and difference with previous work on distance estimation in real and virtual environments.

The goal of our work is to develop a programmatically controlled peer to ride with a human subject for the purpose of studying how social interactions influence riding behavior. The peer is controlled through a combination of reactive controllers that determine the gross motion of the virtual bicycle, action-based controllers that animate the virtual bicyclist and generate verbal behaviors, and a keyboard interface that allows an experimenter to initiate the virtual bicyclist's actions during the course of an experiment. The virtual bicyclist's repertoire of behaviors includes road following, riding alongside the human rider, stopping at intersections, and crossing intersections through specified gaps. The virtual cyclist engages the human subject through gaze, gesture, and verbal interactions. We describe the structure of the behavior code and report the results of a pilot study examining how 10- and 12-year-old children interact with a peer cyclist. Results of the pilot study showed that the presence of the peer had a significant influence on the size of the gaps taken as well as time left to spare between the participant and the trailing car in the crossed gap.

This paper presents the Environment Description Framework (EDF) for modeling complex networks of intersecting roads and pathways in virtual environments. EDF represents information about the layout of streets and sidewalks, the rules that govern behavior on roads and walkways, and the locations of agents with respect to navigable structures. The framework serves as the substrate on which behavior programs for autonomous vehicles and pedestrians are built. Pathways are modeled as ribbons in space. The ribbon structure provides a natural coordinate frame for defining the local geometry of navigable surfaces. EDF includes a powerful runtime interface supported by robust and efficient code for locating objects on the ribbon network, for mapping between Cartesian and ribbon coordinates, and for determining behavioral constraints imposed by the environment.

This paper presents steering behaviors that control autonomous vehicles populating roadways in virtual urban environments. Behavior programming is facilitated by a set of representations of the environment that use convenient frames of reference in natural coordinate systems. Roadway surfaces are modeled as three-dimensional ribbons that make the local orientation of the road explicit and allow relative distances on the road to be simply computed. Roads and intersections are connected to form a ribbon network. An egocentric representation called a path melds road and intersection segments into a single, continuous ribbon that captures the vehicle’s shortterm plan of navigation. A topological structure called a route supports wayfinding. We describe how the interrelated ribbon, path, and route representations are used to build multi-component behaviors that plan routes and safely navigate through traffic filled road networks – tracking lanes, shifting lanes to avoid congestion, anticipating lane changes needed to make turns dictated by the route, negotiating intersections, and respecting the rules of the road.

We conducted three experiments to compare distance perception in real and virtual environments. In Experiment 1, adults estimated how long it would take to walk to targets in real and virtual environments by starting and stopping a stopwatch while looking at a target person standing between 20 and 120 ft away. The real environment was a large grassy lawn in front of a university building. We replicated this scene in our virtual environment using a nonstereoscopic, large-screen immersive display system. We found that people underestimated time to walk in both environments for distances of 40 to 60 ft and beyond. However, time-to-walk estimates were virtually identical across the two environments, particularly when people made real environment estimates first. In Experiment 2, 10- and 12-year-old children and adults estimated time to walk in real and virtual environments both with and without vision. Adults underestimated time to walk in both environments for distances of 60 to 80 ft and beyond. Again, their estimates were virtually identical in the real and virtual environment both with and without vision. Twelve-year-olds' time-to-walk estimates were also very similar across the two environments under both viewing conditions, but 10-year-olds exhibited greater underestimation in the virtual than in the real environment. A third experiment showed that adults' time-to-walk estimates were virtually identical to walking without vision. We conclude that distance perception may be better in virtual environments involving large-screen immersive displays than in those involving head-mounted displays (HMDS).

The paper presents a real-time database for modeling complex networks of intersecting roads and walkways in urban virtual environments. The database represents information about the layout of streets and sidewalks, the rules that govern behavior on roads and walkways, and the locations of agents with respect to road and sidewalk structures. This information is used by programs that control the behavior of autonomous vehicles and pedestrians populating the virtual urban environment. Roads and sidewalks are modeled as ribbons in space. The ribbon structure provides a natural coordinate frame for defining the local geometry of navigable surfaces. This geometry is important for way finding and also forms the geometric basis on which spatial relationships among agents are defined. The database includes a powerful run-time interface supported by robust and efficient code for locating objects on the ribbon network, for mapping between Cartesian and ribbon coordinates, and for determining behavioral constraints imposed by the environment.

This paper presents algorithmic methods to generate progressive, whip-type motions, characteristic of experienced humans in high speed athletic activities like throwing and striking. In deriving these methods, we deduce boundary conditions for dynamic quantities by analysing the motion of a two-link system. A control algorithm based on cascading gains is introduced and is shown to be efficient when compared to fixed-gain proportional control. Principles of energy redistribution within links are used to explain the mechanical advantage gained by a whip-type motion, and the efficiency of cascading gain control. The results can be applied in motion synthesis for animation which is demonstrated through simulation of a multi-link system. Our aim has been to obtain biomechanically plausible solutions which approximate natural motions.

Abstract This paper examines the control of complex physical objects in simulation. We introduce a programming paradigm that allows a simulation to be treated as a multi‐level constraint solver. The control programmer is given the ability to specify constraints on the controlled response of mechanisms and to conditionally change these constraints dependent on the state of system. The approach facilitates the development of model‐based, event‐driven control programs. The usefulness of the paradigm is demonstrated through the simulation of a hopping robot.

A model-independent method for controlling a hopping robot is presented. The approach focuses on the interaction between the hopper and its surroundings at points of contact. It is only through reactions at contact points that a hopper can alter its momenta. A generalization of the virtual leg abstraction allows control programs to be expressed as constraints on the external forces and torques acting on the hopper. The relationship between the nature of the external contacts and motion control strategies is investigated. The approach was tested in simulation using the model-driven dynamic simulator Newton. The experiments demonstrate the utility of simulation for studying physical control problems. Through simulation, a broad range of design parameters can be easily tested to determine the limits of the control strategy.>

A method to estimate the intrinsic and extrinsic patterns of a camera model is presented. The intrinsic parameters of camera center and focal length are estimated with a calibration device which is adjusted iteratively with four independent motions. Experimentation demonstrates that inexperienced users rapidly converge to satisfactory estimates. Extrinsic parameters define the position and orientation of the camera in a world coordinate frame. The straightforward formulation of this mapping leads to a complex system of nonlinear equations. Solutions based on nonlinear estimation techniques used unconstrained sets of known target points. These methods may converge to an incorrect solution and are characteristically ill-conditioned. By imposing a simple regularity on the arrangement of calibration points, the problem can be decomposed into a series of simple one- or two-parameter linear equations. The approach is described, and empirical tests of the accuracy of the method are given.>

Multiple views of a scene can provide important information about the structure and dynamic behavior of three-dimensional objects. Many of the methods that recover this information require the determination of optical flow-the velocity, on the image, of visible points on object surfaces. An important class of techniques for estimating optical flow depend on the relationship between the gradients of image brightness. While gradient-based methods have been widely studied, little attention has been paid to accuracy and reliability of the approach. Gradient-based methods are sensitive to conditions commonly encountered in real imagery. Highly textured surfaces, large areas of constant brightness, motion boundaries, and depth discontinuities can all be troublesome for gradient-based methods. Fortunately, these problematic areas are usually localized can be identified in the image. In this paper we examine the sources of errors for gradient-based techniques that locally solve for optical flow. These methods assume that optical flow is constant in a small neighborhood. The consequence of violating in this assumption is examined. The causes of measurement errors and the determinants of the conditioning of the solution system are also considered. By understanding how errors arise, we are able to define the inherent limitations of the technique, obtain estimates of the accuracy of computed values, enhance the performance of the technique, and demonstrate the informative value of some types of error.