Human Interaction with Assistive Free-Flyers
Transcription
Human Interaction with Assistive Free-Flyers
Human Interaction with Assistive Free-Flyers Daniel J. Szafir Department of Computer Sciences University of Wisconsin–Madison 1210 West Dayton Street Madison, WI 53706 USA dszafir@cs.wisc.edu ABSTRACT Free-flying robots represent a novel platform that appears uniquely suited to assist humans in exploratory, surveillance, inspection, and telepresence tasks across a variety of domains. Such tasks will require “assistive free-flyers” (AFFs), embodied as micro air vehicles (MAVs), robotic blimps, or space robots, to effectively interact with humans in close proximity. I propose an investigation into the design space of proximal AFF interactions, specifically by examining how AFFs might effectively communicate with colocated humans as well as gain an understanding regarding the ecological fit of AFFs within human workspaces. This research takes a two-phase approach towards examining proximal AFF interactions: (1) examining AFF communication mechanisms including motion, body language, and electronic signals, and (2) developing an understanding of user mental models and social expectations for flying robots, including an examination of spontaneous AFF interactions. My work will inform the design of future AFF systems and aid in understandings regarding how AFFs can reach their potential as collaborators within human environments. 1. INTRODUCTION Small free-flying robots hold great potential due to their unique abilities to freely traverse and survey environments. We term these robots, represented in the growing body of research on micro air vehicles (MAVs), robotic airships, and assistive space robots, “assistive free-flyers” (AFFs). AFFs are envisioned to provide aid in domains including construction, utilities, search-and-rescue, and space by performing inspection, mapping, telepresence, and delivery tasks. Due to the assistive nature of these tasks, AFFs will require much greater collaboration and colocated interaction with humans. For example, NASA is currently developing an AFF called “Smart SPHERES,” which is designed to be used inside spacecraft (e.g., the International Space Station) when humans are present to perform a variety of mobile sensing tasks including environmental monitoring (e.g., air quality, radiation, sound levels), autonomous logistics management (e.g., inventory), and mobile camera work to support astronaut activities and payload experiments [3]. Designers must account for human perceptions of AFFs to enable flyers such as the Smart SPHERES to successfully integrate into workflows in human environments. My research explores how AFFs might better communicate their state, intentions, capabilities, tasks, and roles to nearby peers, how local users will relate to flying robots, and how to use these understandings to improve collaborative outcomes for proximal AFF interactions. Figure 1: Assistive free-flying robots are increasingly prevalent in human environments (pictured: NASA SPHERES, Parrot AR Drone 2.0, Autonomous Light Air Vessel, DJI Phantom, Project Skye). This work proposes an in-depth exploration of the many factors influencing proxemic human-free-flyer. 2. RESEARCH PLAN My research uses two phases to analyze proximal human interactions with AFFs. The first phase consists of a series of studies to examine how AFFs can effectively communicate with colocated peers using mechanisms including flight motions, embodied gestures, and electronic signals. In the second phase, I analyze local users’ perceptions and responses to proximal AFFs by examining expected social norms for flying robots and instinctual responses towards AFFs working in human environments. 2.1 Communication for Proximal Interactions AFFs must communicate their intentions, state, and roles to colocated peers to effectively work and collaborate. Misunderstandings regarding AFF communication may damage human-robot rapport, prove detrimental to task efficiency, and may even be dangerous for the human collaborator. The goal of this phase is to examine the communicative affordances of several AFF communication channels within the context of proximal AFF interactions. Study 1: Flight and Intent (Completed): In this study [7], I sought to explore how designers might improve AFF motions to more effectively communicate intent. I first analyzed the motion design space by deconstructing high-level AFF flight paths into composite primitives, such as hovering, approaching a person, or departing an object. Drawing inspiration from computer animation research on creating natural motion, I developed several parameterized manipulations to modify the motions of these primitives, including using arced trajectories, easing AFF velocities smoothly in and out, and using anticipatory motions to telegraph AFF expectations and users’ basic “interaction instincts.” Study 1: Social Norms for Flying Robots (Planned): Currently, AFFs rate low across all dimensions of the social robot framework [2]. In this study, I plan to examine user expectations regarding appropriate social norms for flying robots. This study will examine several aspects including proxemic distances, altitudes, morphology, directionality, and orientation, to gain an understanding regarding the “correct” social behavior for AFFs given that many AFF capabilities do not appear to map directly into any known human social norms. This study will utilize several methods including think-alouds, interviews, and evaluations to gain insight into the design of appropriate AFF behaviors. Figure 2: Observers view a colocated AFF in an environmental inspection task. intent. I then conducted a formative exploration using a virtual AFF to investigated the usefulness of these manipulations in conveying AFF intent. My findings informed the design of manipulations to primitive motions for a physically embodied AFF to use while executing task-based flight paths. I conducted an in-person 2 × 1 within-participants study to evaluate participant responses when observing a colocated AFF executing flight paths constructed from primitives designed to express intent and paths constructed from unaltered primitives. My results showed that the manipulated motion designs significantly improved viewers’ preferences for working with an AFF, ratings of motion naturalness, and sense of safety. Study 2: Signalling Mechanisms for AFFs (Ongoing): Although there has been a great deal of research on nonverbal robot communication (e.g., see [4]), it is not always clear how to adapt findings to functional robots that lack zoomorphic or anthropomorphic features. This study will examine the development of a signalling “language” for AFFs, using three channels: luminescence, sound, and motion. Each channel is able to express cues by manipulating a variety of variables, for instance manipulating the colors, intensities, and positions of electronic lights. I first plan to develop a comprehensive library of potential AFF signalling mechanisms from prior work (e.g., [1, 5]) and map the potential affordances of these mechanisms to various AFF communication goals. I then plan to evaluate the design using a 3 (correctly mapped vs randomly mapped vs no signalling mechanism) x 2 (collaborator vs bystander) study in which participants interact with an embodied AFF executing an environmental inspection task (e.g., Figure 2). My hypothesis is that the utilization of signalling mechanisms affording the correct communication goal will maximize perceptions of AFF usability and trust and will increase task performance for proximal collaborators. 2.2 Mental Models and Expectations Responses to robot aid are extremely complex [8], often due to unclear mental models of how users should interact with the robot or interactions in which the robot violates social norms or expectations. This phase will examine of how people make sense of flying robots by examining AFF social Study 2: Spontaneous Interactions (Planned): AFFs are envisioned to interact extensively within human environments where even bystanders may exhibit limited interaction [6]. Examining bystander interactions, especially interactions which may change a bystander into a peer, hold great promise for increasing AFF usability. In this study, I plan to examine limited, spontaneous bystander interactions with AFFs to analyze both the form of these interactions (e.g., gaze, gestures) as well as bystander goals during these interactions (e.g., indicate low interruptibility, show interest in the robot). My hypothesis is that AFFs that can understand and recognize the instinctual interaction methods used by bystanders in spontaneous interaction will be seen as more useful and less intrusive in human work environments. 3. EXPECTED CONTRIBUTIONS This research will produce several contributions the body of HRI knowledge by helping to develop a framework for AFF communicative affordances and grounding future studies with an exploration of how users develop mental models and social understandings for interacting with AFFs in close proximity. This research will produce an increased contextual understanding surrounding proximal human interactions with AFFs, which are currently a novel and extremely promising platform for human-robot collaboration. 4. REFERENCES [1] S. Andrist, T. Pejsa, B. Mutlu, and M. Gleicher. Designing effective gaze mechanisms for virtual agents. In Proc CHI’12, pages 705–714, 2012. [2] C. Bartneck and J. Forlizzi. A design-centred framework for social human-robot interaction. In Proc ROMAN’04, pages 591–594, 2004. [3] T. Fong, M. Micire, T. Morse, E. Park, C. Provencher, V. To, D. Wheeler, D. Mittman, R. J. Torres, and E. Smith. Smart spheres: a telerobotic free-flyer for intravehicular activities in space. In Proc. AIAA Space’13, 2013. [4] T. Fong, I. Nourbakhsh, and K. Dautenhahn. A survey of socially interactive robots. Robotics and Autonomous Systems, 42(3–4):143–166, 2003. [5] C. Harrison, J. Horstman, G. Hsieh, and S. Hudson. Unlocking the expressivity of point lights. In Proc CHI’12, pages 1683–1692, 2012. [6] J. Scholtz. Theory and evaluation of human robot interactions. In Proc System Sciences’03, 2003. [7] D. Szafir and B. Mutlu. Communication of intent in assistive free flyers. In Proc HRI’14 (To Appear), 2014. [8] C. Torrey, S. Adviser-Kiesler, and S. R. Adviser-Fussell. How robots can help: Communication strategies that improve social outcomes. 2009.