Visuo-Haptic Collaborative Augmented Reality Ping-Pong

Page created by Rodney Turner
 
CONTINUE READING
Visuo-Haptic Collaborative Augmented Reality Ping-Pong
Visuo-Haptic Collaborative Augmented Reality Ping-Pong

                        B. Knoerlein                                        G. Székely                         M. Harders
              knoerlein@vision.ee.ethz.ch                        szekely@vision.ee.ethz.ch              harders@vision.ee.ethz.ch
                                                                Computer Vision Laboratory
                                                                       ETH Zurich
                                                                CH-8092 Zurich, Switzerland

ABSTRACT                                                                             interfaces are necessary, which allow transferring skills and
In our current work we examine the development of visuo-                             knowledge from and to daily life.
haptic augmented reality setups and their extension to col-                             Using augmented reality environments as user interfaces
laborative experiences in entertainment settings. To this                            for games has been a topic of research for a while, as de-
end, an expandable system architecture supporting mul-                               scribed for instance in [7, 12, 18, 5, 8]. Some research re-
tiple users is one of the most indispensable prerequisites.                          lated directly to our example application of AR ping-pong
In addition, system stability, low latency, accurate calibra-                        have been carried out in the past. For instance, in [19] a
tion and stable overlay of the virtual objects have to be                            virtual table-tennis setup with real rackets has been devel-
assured. In this paper we provide an overview of our frame-                          oped. However, the system did not include haptic feedback
work and present our collaborative example application, an                           and only used the rackets as an interface to position a vir-
augmented reality visuo-haptic ping-pong game for two play-                          tual racket on the screen. A similar approach was also taken
ers. The users play with a virtual ball in a real environment                        in [11], which neither provided haptic feedback. In [13], a
while, by using virtual bats colocated with haptic devices,                          collaborative, mobile AR-based billiard game has been dis-
they are able to feel the impact of the simulated ball on the                        cussed, already incorporating tactile feedback. In most of
bat.                                                                                 these works, either no details on overlay accuracy, stabil-
                                                                                     ity and latency are given, or the reported results are of
                                                                                     lower quality. Regarding the integration of haptic feedback
Categories and Subject Descriptors                                                   in general AR setups only limited work has been carried
H.5.1 [Multimedia Information Systems]: Artificial, aug-                             out. Some groups have experimented with adding props
mented, and virtual realities; I.3.7 [Three-Dimensional                              to provide passive haptic feedback in the AR environment,
Graphics and Realism]: Virtual reality                                               e.g.[14, 6, 15]. However, active behavior of scene objects can
                                                                                     not be generated with these systems. Recent research has
General Terms                                                                        examined the possibility of adding devices providing active
                                                                                     haptic feedback. A few proof-of-concept systems have been
Design, Performance
                                                                                     developed in this context, for instance [17, 10, 20, 1]. How-
                                                                                     ever, lag reduction, exact alignment, or error minimization
Keywords                                                                             is only seldomly addressed. Moreover, the haptic device is
Augmented reality, haptics, collaboration                                            usually not colocated with the visual representation of the
                                                                                     augmented objects, thus compromising hand-eye coordina-
1.     INTRODUCTION                                                                  tion. In addition, if simultaneous interaction with real and
                                                                                     virtual objects should be allowed, visuo-haptic colocation in
   The underlying idea of augmented reality (AR) systems is
                                                                                     the augmented scene is indispensable. In this context, the
the combination of real and virtual objects into one environ-
                                                                                     real world represents a reference frame into which the vir-
ment [2]. However, the perception of the augmented scene
                                                                                     tual elements have to be perfectly integrated. Any errors
is usually restricted to the visual domain and no haptic cues
                                                                                     in this process would reduce the usability of a system, and
are provided, limiting user interactions. The integration of
                                                                                     compromise user interaction and immersion.
haptic interfaces into AR systems removes this constraint
                                                                                        To alleviate the currently existing shortcomings, we have
and permits to not only see the virtual objects, but in ad-
                                                                                     developed a high precision, collaborative, colocated visuo-
dition to feel them. This can be used for numerous appli-
                                                                                     haptic augmented reality environment. Main components
cations where natural, three-dimensional human computer
                                                                                     are a distributed, low latency framework for synchroniza-
                                                                                     tion, the accurate calibration of the haptic device in world
                                                                                     coordinates and a hybrid tracking setup. In this short paper
Permission to make digital or hard copies of all or part of this work for            we describe the developed framework and present an exam-
personal or classroom use is granted without fee provided that copies are            ple testbed – multimodal AR ping pong. Figure 1 depicts
not made or distributed for profit or commercial advantage and that copies           the overall setup for one user as well as his augmented view
bear this notice and the full citation on the first page. To copy otherwise, to      of the scene.
republish, to post on servers or to redistribute to lists, requires prior specific
permission and/or a fee.
ACE’07, June 13–15, 2007, Salzburg, Austria.
Copyright 2007 ACM 978-1-59593-640-0/07/0006 ...$5.00.
Visuo-Haptic Collaborative Augmented Reality Ping-Pong
Graphic loop                                                      Physic                                      Haptic
                                                                                                                                    loop                                        loop
                                                                                                                                                            Transformtations
                                                                 Grabbing of image                  requ                           1 cycle                                     1 cycle
                                                                                                        est                                  Forces
                                                                                                                                                            Transformtations
                                                                                                                                   1 cycle                                     1 cycle
                                                                  Undistortion of                                                            Forces         Transformtations
                                                                                                                                   1 cycle                                     1 cycle
                                                                      image                                                                  Forces         Transformtations
                                                                                                                                   1 cycle   Forces
                                                                                                                                                                               1 cycle
                                                                  Pose refinement                                       sea                                 Transformtations
                                                                                                              ans           r      1 cycle                                     1 cycle
                                                                                                                 we
                                                                                                                    r   pa ching   1 cycle
                                                                                                                                             Forces         Transformtations
                                                                                                                                                                               1 cycle
                                                                    Drawing of                                            cke                Forces
                                                                                                                              t
                                                                    background                g                                    1 cycle                                     1 cycle

                                                                                                                                                      ...
                                                                                          ckin te
                                                                    Drawing of         che upda                                    1 cycle                                     1 cycle
                                                                   static objects        for                                       1 cycle                                     1 cycle
                                                                                                                                   1 cycle                                     1 cycle
                                                                     Drawing of                                                    1 cycle                                     1 cycle
                                                                 interactive objects                     st
                                                                                                    reque                          1 cycle                                     1 cycle
                                                                 Grabbing of image

                                                                                                                                    ...

                                                                                                                                                                                ...
                                                                        ...
                                                                 Figure 2: Data exchange in the distributed system.

                                                                 nected. A second haptic device is attached to a third 1.0GHz
                                                                 single CPU machine (256MB RAM, 256kb cache) running a
                                                                 haptics client. All components are using Linux OS. As ex-
Figure 1: View of the colocated visuo-haptic AR                  ternal tracker we use the infrared (IR) optical 6DoF tracking
system.                                                          device OPTOTRAK 3020 manufactured by Northern Digi-
                                                                 tal Inc. The view of one user is captured by a head-mounted
                                                                 Videre Design MEGA-DCS FireWire camera (640x480 pixel
2.   COLLABORATIVE AR FRAMEWORK                                  image at 30 fps). In addition, a user-independent top view
   The computational requirements for achieving stable mul-      of the scene is generated by an Allied Vision Technologies
timodal interaction in multiuser AR go beyond the capabil-       FireWire camera (800x600 at 30fps). Since only one video-
ities of currently available hardware. Therefore, the sys-       see-through HMD was available, one player currently has to
tem framework had to be distributed to several machines.         rely on the images displayed via the overhead camera. A
The core components of our system are the physics server, a      second HMD setup is already manufactured for our system.
graphics and a haptics client. The first machine performs the    Finally, to allow touching of virtual objects, two SensAble
physical simulation of the virtual scene. A graphics client      1.5 PHANToM devices are integrated into our AR setup.
carries out all tasks typical to a standard AR setup. The        As previously mentioned, the haptic devices are colocated
external input data needed on this client are position and       with the augmented environment, thus allowing interaction
orientation of the haptic interfaces, as well as the move-       with virtual and real objects via the same tool. Therefore,
ment of the virtual objects in the scene. Finally, haptic        high demands for tracking and calibration have to be met,
interfaces can be added to the environment with the haptics      which will be discussed in the following.
clients, which communicate directly with the physics server.
In our example setup, one device is directly connected to
the physics server, while a second interface is added via a      3. HYBRID TRACKING
haptics client. To provide high level of immersion, the vi-        An IR marker is attached to the head-mounted camera
sual data acquired from the simulation always has to coin-       and tracked by the OPTOTRAK system to determine the
cide with the moment the image was taken on the graphics-        user’s head pose. The system measures and computes marker
machine. Therefore, accurate synchronization between the         orientation and position with submillimeter accuracy at an
physics server and a graphics client is required. To this        optimal distance of about two meters. Since the camera and
end, the actual clock difference is determined at startup of     the marker are rigidly attached, the camera-marker trans-
the application. This computation is done by a comparison        formation is fixed and can thus be determined by Hand-Eye
of the clocks on both machines, while taking into account        calibration [16]. The IR tracking data are inherently noisy
the round trip time. After synchronizing both clocks, the        due to the inaccurate measurements of the LED positions,
graphical data is stored in a timestamped ringbuffer and the     especially when the head-mounted marker moves. As a con-
graphics client can request data for the timestamp the im-       sequence, the registration between the real and the virtual
age was taken. The data are transmitted, temporarily stored      world is affected, causing instability of virtual objects in
and updated when needed by the rendering process. There-         the augmented images. Therefore, we correct the estimated
fore, no latency is added to the overall delay. One round trip   camera pose of the IR optical tracker with a vision-based
of the communication takes 500−900µs for small-sized pack-       approach. The first step is to detect precalibrated visual
ets. The data exchange between the physics server and the        landmarks in the recorded images. Furthermore, to ensure
haptics client is collision free with a round trip time below    that a landmark is sufficiently visible, an occlusion test is
200µs. Figure 2 illustrates all information communication        carried out and occluded landmarks are discarded. Given
occurring in the framework.                                      the 2D-3D correspondence of the remaining landmarks, the
   In our testbed setup for two-player interaction in multi-     position and the orientation of the camera are then refined
modal AR, the physics server is run on a dual processor PC       by error minimization. An image space approach is used to
with 2.4GHz CPUs (2GB RAM, 512kb cache). Rendering               estimate the camera pose as a nonlinear least squares prob-
and tracking on the graphics client is done on a dual pro-       lem. A RMS backprojection error of less than 0.7 pixels can
cessor PC with 2.8GHz CPUs (2GB RAM, 512kb cache).               be achieved in less than 1ms computation time. Moreover,
To the former a PHANToM is attached, while to the latter         using additional offline refinement of 3D landmark positions,
the optical position tracking device and the cameras are con-    the error can be further reduced to about 0.3 pixels [3].
Visuo-Haptic Collaborative Augmented Reality Ping-Pong
pivot calibration [9], giving the marker-tip transformation.
                                                                 This allows obtaining corresponding point measurements in
                                                                 both the haptic and world coordinate system. After acquir-
                                                                 ing 3D point data, the absolute orientation problem has to be
                                                                 solved. Due to additional errors, resulting from inaccuracies
                                                                 in haptic encoder initialization, a two-staged optimization
                                                                 process is followed [4]. The final calibration results yield an
                                                                 alignment error better than 1.5mm. Figure 3 demonstrates
                                                                 the calibration results for visuo-haptic colocation. A virtual
                                                                 table tennis bat is aligned with a real handle attached to the
                                                                 haptic device.

                                                                 5. COLLABORATIVE, VISUO-HAPTIC AR
                                                                    PING-PONG
                                                                    To demonstrate the high fidelity of the system, an AR
                                                                 ping-pong environment was chosen as testbed scenario for
                                                                 our collaborative visuo-haptic augmented reality system. This
                                                                 choice has been driven by the importance of fast collabo-
                                                                 rative interactions during gameplay. Real handles of table
                                                                 tennis rackets are attached to the haptic devices. The actual
                                                                 bat and the ball are, however, overlaid virtual objects. In
                                                                 order to integrate the real table and the net, these objects
                                                                 are initially calibrated in the world coordinate system, thus,
                                                                 allowing the virtual ball to collide with them. A simplified
                                                                 rigid body dynamics model is used for physical modeling
                                                                 of the ball. The haptic devices are used to render impact
                                                                 forces to the users and to control the virtual bat. The overall
                                                                 rendering performance of the augmented scene at a resolu-
                                                                 tion of 640x480 pixels is illustrated in Table 1. Due to the
                                                                 multi-threaded implementation, the images are displayed at
                                                                 about 20Hz with an overall latency of about 60 − 100ms.
                                                                 The scene as shown to one of the users is depicted in Figure
                                                                 4.

                                                                           Component                     Average time [ms]
                                                                          Frame rendering                       15
                                                                         Frame unwrapping                        5
                                                                  Corner detection + occlusion test             20
                                                                          Pose refinement                       0.6
                                                                          Object rendering                       5

                                                                            Table 1: Rendering performance

Figure 3: Haptic device without (top) and with colo-
cated augmentation of a virtual bat (center,bottom) .            6. CONCLUSION
                                                                    We have presented a colocated visuo-haptic augmented
                                                                 reality environment for two-player ping-pong. A dedicated
4.   HAPTIC DEVICE CALIBRATION                                   system providing accurate calibration, high stability, and
   In order to align the virtual representation of the hap-      low latency had to be developed. A problem of the current
tic interaction point with the correct physical location in      setup is the lack of binocular stereo cues. This limits the
the real world, the relationship between the haptic and the      interactivity of the system since depth information has to
world coordinate system needs to be determined. The first        be inferred from shadows, occlusions and motion parallax.
step of our calibration procedure is to collect 3D point mea-    Moreover, the workspace of the currently used haptic device
surements in the coordinate systems of the haptic device and     is rather limited. Therefore interactions are restricted to a
the optical tracker. To this end an optical marker is attached   certain area. In addition, due to the used device only 3DOF
to the stylus of a PHANToM device. For the measurements,         of force feedback can be provided.
the tip position is stabilized in the middle of the haptic          Next steps will be the extension to stereo vision for better
workspace by rendering a fixation force with the device. The     depth perception. In addition, to reduce network traffic and
marker assembly is then manually rotated following a spher-      to ease the synchronization a hardware trigger signal will be
ical movement in space, while the marker poses are being         integrated. Also, extension to 6DOF haptic feedback and
recorded. The best fitting sphere is then determined using       alternative haptic devices will be investigated.
Visuo-Haptic Collaborative Augmented Reality Ping-Pong
and I. Rudomin. Multi-user networked interactive
                                                                       augmented reality card game. In International
                                                                       Conference on Cyberworlds, pages 177–182, 2006.
                                                                 [6]   R. Grasset, J.-D. Gascuel, and D. Schmalstieg.
                                                                       Interactive mediated reality. In Symposium on Mixed
                                                                       and Augmented Reality, pages 302–303, 2003.
                                                                 [7]   T. Jebara, C. Eyster, J. Weaver, T. Starner, and
                                                                       A. Pentland. Stochasticks: Augmenting the billiards
                                                                       experience with probabilistic vision and wearable
                                                                       computers. In Proc. 1st Int’l Symp. Wearable
                                                                       Computers ISWC, pages 138–145, 1997.
                                                                 [8]   A. Lam, K. Chow, E. Yau, and M. Lyu. Art:
                                                                       augmented reality table for interactive trading card
                                                                       game. In Proceedings of the 2006 ACM international
                                                                       conference on Virtual reality continuum and its
                                                                       applications, pages 357–360, 2006.
                                                                 [9]   S. Lavalle, P. Cinquin, and J. Troccaz. Computer
                                                                       Integrated Surgery and Therapy: State of the Art,
                                                                       chapter 10, pages 239–310. IOS Press, 1997.
                                                                [10]   T. Nojima, D. Sekiguchi, M. Inami, and S. Tachi. The
                                                                       smarttool: a system for augmented reality of haptics.
                                                                       In Proc. of IEEE Virtual Reality, 2002, pages 67–72,
                                                                       2002.
                                                                [11]   T. Ohshima, K. Satoh, H. Yamamoto, and H. Tamura.
                                                                       AR2 hockey: A case study of collaborative augmented
                                                                       reality. In Proc. VRAIS, pages 268–275, 1998.
                                                                [12]   W. Piekarski and B. Thomas. Arquake: The outdoor
                                                                       augmented reality gaming system. Communications of
                                                                       the ACM, 45(1):36–38, 2002.
                                                                [13]   U. Sargaana, H. S. Farahani, J. W. Lee, J. Ryu, and
                                                                       W. Woo. Collaborative billiARds: Towards the
                                                                       ultimate gaming experience. In ICEC, pages 357–367,
                                                                       2005.
                                                                [14]   D. Schmalstieg, A. Fuhrmann, G. Hesina, Z. Szalavari,
                                                                       L. M. Encarnacao, M. Gervautz, and W. Purgathofer.
Figure 4: Colocated visuo-haptic AR ping-pong for
                                                                       The studierstube augmented reality project. Presence
two players.
                                                                       - Teleoperators and Virtual Environments,
                                                                       11(1):32–45, 2002.
7.   ACKNOWLEDGMENTS                                            [15]   R. Sidharta, J. Oliver, and A. Sannier. Augmented
                                                                       reality tangible interface for distributed design review.
   This work has been performed within the frame of the
                                                                       In Computer Graphics, Imaging and Visualisation,
Swiss National Center of Competence in Research on Com-
                                                                       pages 464–470, 2006.
puter Aided and Image Guided Medical Interventions (NCCR
Co-Me) supported by the Swiss National Science Founda-          [16]   R. Y. Tsai and R. K. Lenz. A new technique for fully
tion.                                                                  autonomous and efficient 3D robotics hand/eye
                                                                       calibration. IEEE Journal of Robotics and
                                                                       Automation, 5(3):345–358, 1989.
8.   REFERENCES                                                 [17]   J. R. Vallino and C. M. Brown. Haptics in augmented
 [1] M. Adcock, M. Hutchins, and C. Gunn. Augmented                    reality. In ICMCS, Vol. 1, pages 195–200, 1999.
     reality haptics: Using ar toolkit for display of haptic    [18]   D. Wagner, T. Pintaric, F. Ledermann, and
     applications. In Proc. of 2nd IEEE Intl Augmented                 D. Schmalstieg. Towards massively multi-user
     Reality Toolkit Workshop, pages 1–2, 2003.                        augmented reality on handheld devices. In Pervasive,
 [2] R. T. Azuma. A survey of augmented reality.                       pages 208–219, 2005.
     Presence: Teleoperators and Virtual Environments,          [19]   C. Woodward, P. Honkamaa, J. Jäppinen, and E.-P.
     6(4):355–385, 1997.                                               Pyökkimies. Camball: augmented networked table
 [3] G. Bianchi, C. Jung, B. Knoerlein, M. Harders, and                tennis played with real rackets. In Advances in Comp.
     G. Székely. High-fidelity visuo-haptic interaction with          Entertainment Techn., pages 275–276, 2004.
     virtual objects in multi-modal ar systems. In ISMAR        [20]   G. Ye, J. J. Corso, G. D. Hager, and A. M. Okamura.
     2006, October 2006.                                               Vishap: Augmented reality combining haptics and
 [4] G. Bianchi, B. Knoerlein, G. Székely, and M. Harders.            vision. In Proc. of IEEE Intl. Conference on Systems,
     High precision augmented reality haptics. In                      Man and Cybernetics, pages 3425–3431, 2003.
     Eurohaptics 2006, pages 169–177, July 2006.
 [5] M. Diaz, M. Alencastre-Miranda, L. Munoz-Gomez,
Visuo-Haptic Collaborative Augmented Reality Ping-Pong Visuo-Haptic Collaborative Augmented Reality Ping-Pong
You can also read