Haptic Battle Pong:

High-Degree-of-Freedom Haptics in a Multiplayer Gaming Environment

Dan Morris
Computer Science Department
Stanford University
[email protected]

Neel Joshi
Computer Science Department
Univ of California, San Diego
[email protected]

Abstract
Recent advances in both haptic rendering techniques and CPU
power have made haptic feedback increasingly accessible on
consumer hardware. With the recent success of simple haptic
feedback devices in console and PC game systems, more
sophisticated haptic rendering is poised to enter the mass
entertainment market in the near future. With this in mind, we
describe Haptic Battle Pong, a competitive networked game that
makes extensive use of three-degree-of-freedom force-feedback
and six-degree-of-freedom input. This is among the first games to
fully utilize the capabilities of high-fidelity haptic devices, and we
use this environment to explore the applicability of high-degreeof-freedom haptics to games.
We discuss our approaches to physical simulation, forcerendering, and networked haptic interaction. We also discuss
several techniques we used to enhance the entertainment value of
a traditional video game using the capabilities of a haptic display.
Finally, we review the subjective responses of initial users and
several lessons we learned that may be applied to future haptic
games.

interfaces. Consumer haptic devices from rumble feedback

devices to more sophisticated two-degree-of-freedom joysticks
have gained significant popularity, and are supported by
numerous games and development APIs (e.g. DirectX).
Given the growing flexibility of high-performance haptics, the
magnitude of recent growth in the video game industry, and the
game industrys initial successes with haptic feedback, it seems
almost inevitable that sophisticated haptics will find a key niche
in electronic gaming.
However, there are still significant engineering and human
interface challenges associated with haptics that have not been
explored by existing video games. This paper describes Haptic
Battle Pong, a demonstration of the potential for haptics in video
games. The project leverages the familiarity of classic Pong, but
also addresses some of the key challenges presented by haptic
gaming. In the context of the game and its development
framework, we address three primary topics:

Tight, closed-loop integration of haptic feedback and

simulation physics. This has not been possible with the
simple haptic devices used in current mass-market games;
commercial games currently use haptics only to transmit
information to the player.

Networked competition between players. Networked haptic

feedback presents numerous challenges associated with the
high update rates required for haptic interfaces and the high
latencies associated with network transmission.

Entertainment value of haptics. It remains to be proven that

high-degree-of-freedom haptics can enhance a players video
game experience; we present some novel techniques for
enhancing traditional gaming environments and initial
responses from users.

1. Introduction
Recent years have seen tremendous growth in the capabilities of
high-performance, multi-degree-of-freedom commercial haptic
devices, such as the SensAble Phantom [Massie and Salisbury
1994) and the Force Dimension Delta [Grange et al 2001].
Although these devices have been largely restricted to research
users and specialty customers, high-end haptic systems have
found applications in a variety of disciplines, including medicine
[Agus et al 2003, Mendoza and Laugier 2003], assistive
technology [Morris and Joshi 2003, Sjstrm 2001], and product
design [Nahvi et al 1998, Zhang et al 2003].
The past few years have also seen an explosion in the popularity
of computer gaming, an industry that totaled over ten billion
dollars in US sales alone in 2002 [Rocsearch, 2003]. With the
advent of powerful CPUs on home computers and game consoles,
the video game industry has begun to expand into multimodal

Kenneth Salisbury
Computer Science Department
Stanford University
[email protected]

2. Game Description
Haptic Battle Pong (HBP) is a two-player networked game; each
player is seated at a computer with a Phantom haptic interface
(SensAble, inc.). We have successfully tested HBP with the
Desktop, 1.0, 1.5, and Omni Phantom models. Our typical
configuration uses a Desktop Phantom and a 1.5GHz Pentium 4.
Graphics are rendered using OpenGL; a stereo display (using
OpenGLs support for stereo rendering) is optional.
The core gameplay is consistent with the traditional Pong game;
each players primary objective is to make contact with a moving
ball using a virtual paddle. The playing area is enclosed by walls
on all sides. Each player is represented by an on-screen avatar,
who can be moved in two dimensions using the keyboard. This

Page 1 of 5

FIGURE 1: A screen shot captured from Haptic Battle Pong. Each avatar represents the center of a players haptic workspace; players
control tools in six degrees of freedom within that workspace.
avatar represents the center of the players workspace; movement
and rotation of the paddle within this workspace are performed
using the Phantom. FIGURE 1 displays a screenshot of HBP,
including the playing area and the player and opponent avatars.
In the competitive, networked mode of HBP, a players territory
includes half of the court; a player can only manipulate the ball
within his or her territory. This simplification which we refer to
as the private haptic space assumption greatly reduces the
complexity of the networking (since haptic interaction is never
precisely simultaneous), and is consistent with the gaming model
of pong-type games and most existing sports-based video
games. As we discuss later, this does not eliminate the sense of
physical interaction among players.
A players objective is to keep the ball in the opponents territory;
a player loses health points each time the ball bounces within
his or her territory. Players also interact via a series of haptic
mines; a player can enter his opponents territory to place a mine
on the court. If a player steps on his opponents mine, he loses
health points and is subject to one of several haptic effects, which
make use of the Phantoms sophisticated force-rendering
capabilities and enhance the level of interaction between players.

3. Physics and Force-Feedback

The most critical aspects of HBPs use of the haptic device
involve the interaction between a players paddle and the ball.
Typical games that make use of the force-feedback effects
available on consumer force-feedback devices simply render
forces to inform the player about certain game events. In contrast,
HBP defines a tight, closed-loop integration between the physics
of the game and the force-rendering performed using the
Phantom.

The interaction between the ball and the paddle is modeled as a

sphere penetrating a solid disc with a fixed spring constant. The
force exerted on the ball is always perpendicular to the plane of
the paddle and is proportional to the balls penetration through the
plane of the paddle. This leads to a fairly intuitive model of
contact from the players perspective, and provides the basic
physical properties that are expected from a ball-paddle collision.
For example, the velocity with which the ball leaves the paddle is
proportional both to the initial velocity of the ball and the initial
velocity of the paddle (i.e., swinging harder makes the ball
move faster).
The haptic display is used to render the force resulting from
contact between the ball and the paddle. Initially, we set the force
rendered on the Phantom to be proportional to the force exerted
on the ball at all times, in the interest of physical realism and
conservation of virtual energy. Initial results showed that this was
not intuitive from a players perspective, and generally led to very
brief contacts that were difficult to control. We experimented
with a variety of force models, and ultimately decided that the
simplest contact model a constant force that is not dependent on
the velocity of the ball or the penetration distance of the ball
was an effective approach. The perceived effect was similar to the
brief contact established between a tennis ball and a tennis
racquet. The force of the impact was still perceived to vary
with the force exerted on the ball, since longer contact times
resulted in more net force being applied to the Phantom.
The only problem with this approach, initially, was that the
Phantom tended to vibrate when a player tried to pick up a
stationary ball, due to the rapid and repeated initiation and release
of contact with the ball. So at very low penetration distances
(which occur only briefly in typical collisions), we apply a force
to the Phantom that is proportional to the force applied to the ball.

Page 2 of 5

to summarize, the force rendered on the Phantom can be

represented as:
Fhaptic =

kp (Pp Pb)

(for |Pp Pb| <= d)

(for |Pp Pb| > d)

...where Fhaptic is the force applied to the haptic device, kp is the

gain used when rendering a linear spring force, Pb is the xyz
position of the ball, Pp is the xyz position of the balls projection
onto the plane of the paddle, d is the maximum penetration
distance at which linear spring forces are rendered, and c is the
constant force applied when the penetration distance is greater
than d.
We also model the balls rotational acceleration resulting from
motion of the paddle within the plane of the paddle surface during
contact with the ball. The induced rotational acceleration is not
tied to the induced linear acceleration; rotational acceleration is
proportional to the tangential motion of the paddle, computed as:

d = dt ((k r (( pball a ) v p )) k rf )
...where is the rotational velocity of the ball, dt is the time
elapsed between physics computation iterations, kr is a constant
linking paddle velocity to rotational acceleration, pball is the
position of the ball, a is the projection of the ball onto the plane of
the paddle surface, vp is the linear velocity of the paddle, and krf is
a constant governing angular deceleration due to contact friction.
The first term on the right side of this equation induces an angular
acceleration according to the tangential velocity of the paddle, and
the second term introduces an angular deceleration due to contact
friction.
Subsequent contacts between the ball and other objects (walls or
paddles) result in a change in the balls linear velocity due to the
rotational momentum of the ball and friction between the ball and
the contact surface. The perceived effect is similar in magnitude
to the spin applied to a ping-pong ball, which causes visible
deviations in the path of the ball upon surface contacts.
When contact between the ball and a users paddle occurs, any
rotational velocity accumulated by the ball induces forces on the
Phantom that are parallel to the plane of the paddle. This
increases the sense of competitive interaction between players,
since one players spinning motion can induce a force on the
opponents paddle that complicates his interaction with the ball.
The acceleration of a players paddle due to contact with a
spinning ball is computed as:

We also found it necessary to apply a non-linear air friction to

both the linear and rotational velocities of the ball. Due to the
wide range of swinging forces and velocities applied by players,
the ball could reach extremely high velocities in some cases,
which resulted in a loss of ball control by the players. We
experimented with simply increasing the virtual mass of the ball
and maintaining the linear acceleration model, but in this case
small motions of the Phantom which were typical among novice
players failed to adequately accelerate the ball. Therefore, we
found that gameplay was enhanced significantly by placing an
absolute upper threshold on ball velocity and applying air friction,
which was modeled as a constant decrease in the magnitude of
linear and angular velocities of the ball when it was not in contact
with an object.
A player can also use his paddle to launch projectile bullets at his
opponent (see FIGURE 1). This allowed us not only to add variety
to the gameplay, but also to explore the application of 6-dof input
and 3-dof haptic feedback to another popular class of video
games. The gun is controlled like the paddle; the Phantom
moves and rotates the gun within the workspace centered around
the avatar. A keyboard button fires a bullet from the gun, which
initiates a transient, constant force response along the axis of the
gun, away from the bullets path (simulating the recoil of the
gun). Similarly, contact between a bullet and a players avatar
initiates a transient, constant force in the direction in which the
bullet was moving.
The final application of haptic feedback in HBP was perhaps the
most exciting for players who had not had extensive experience
using the haptic devices. Each player is given four haptic
mines, which can be placed in the opponents territory. When a
player steps on a mine left by his opponent, he experiences a
rumbling effect, similar to that provided by a standard forcefeedback game pad. However, he also experiences a temporary
haptic effect that is specific to the mine he encountered. The
effects are:

Slow paddle : The players paddle is subject to a viscous

force, equal to a constant (negative) multiple of the paddles
velocity.

Heavy paddle : The players paddle is subject to a

gravitational force.

Remote operation paddle : The Phantom held by the player

who encountered the mine is connected by a virtual spring to
the Phantom of the player who placed the mine, a simple
form of tele-operation. I.e. if player 2 is subject to this
effect, the force applied to his haptic device is computed as:

f hr = k hr (( R p ) ( pball a ))

fp2 = (pp1 pp2)kf

...where fhr is the force applied to the haptic device due to spin, khr
is a constant linking tangential force applied to the paddle to
haptic force, Rp is the rotation matrix defining the rotation of the
paddle face, is the world-space rotational velocity of the ball,
and pball and a are defined above. Note that conceptually a torque
should additionally be applied to rotate the paddle around the ball,
but most commercially-available haptic devices are unable to
render torque, so we apply only the linear component of induced
forces.

Page 3 of 5

...here fp2 is the force applied to player 2s haptic device, pp2

is the position of player 2s haptic device within its
workspace, and pp1 is the position of player 1s haptic
device within its workspace, mirrored around the plane that
divides the left and right halves of the workspace. This
mirroring allows player 1 to more intuitively influence the
movements of player 2s haptic device, which we found
enhanced the entertainment value of the remote operation
paddle interaction.

The fourth mine is not associated with a haptic effect; it causes

only health damage and a rumbling force.

4. Networking
In order to avoid the latency-dependence experienced by many
networked haptic environments, we make a private haptic space
assumption: only one player can be in physical contact with the
ball at one time, and players never directly come into haptic
contact with each other. The ball is always within the territory of
one player; this player is considered the master, and his
computers physical simulation determines the actual position
of the ball.
We propose that these assumptions will generalize to a wide
variety of games, including most sports-based and shooting-based
games. Note that it does not require that these zones be fixed over
time, only that a single simulation master can be defined at all
points in space at any given time.
These assumptions would not necessarily hold for generalpurpose simulation environments, such as those used in surgical
simulation research, or for games based on concurrent physical
interaction, e.g. fencing or boxing.
Given these assumptions, the networking protocol aims to allow
each computer to run its own physical simulation when the ball is
in contact with the local player, to allow high haptic update rates
during this critical period.
In fact, the networking protocol always allows physical simulation
to run concurrently on both computers. Each simulation owns
the territory that is occupied by the local player. When a
simulation finds that the ball is in its territory, it is temporarily the
simulation master. In this state, HBP will transmit periodic
simulation updates (position and velocity of the ball) to the
remote machine. The remote machine will update its simulation,
and continue with local physics computation until it receives
another update. When the master determines that the ball has
passed into the remote players territory, it sends a final
simulation update to indicate this status change, and the remote
machine becomes the simulation master. At all times, each
computer periodically transmits the position and orientation of the
local player and his paddle.
This approach allows each computer to maintain a highly accurate
local physics model when the ball is in contact with the local
player. When the ball is in contact with the remote player, the
periodic updates are more than adequate for keeping up with the
graphic frame rate (~100Hz), even when the local player is
separated from the remote player by transcontinental distances.
Dropped or delayed packets rarely cause visible glitches, since the
physical simulation is updated on both machines at all times.

Initial responses from players indicate that the haptic aspects of

the game are intriguing and effective, enhancing the sense of
interaction with an opposing player relative to a standard video
game.
However, players particularly players who had limited
experience using the Phantom initially found the six-degree-offreedom input difficult to control. After discussions with players,
we made several changes to the game.
One of the most challenging aspects of making contact with the
ball was the need for very accurate depth perception. To this end,
we added a shadow on the ball which significantly improved
position perception and the ability to use stereo glasses with
HBP when running in full-screen mode.
Initial responses to these changes are positive, although there is
still a significant learning curve for novice users, during which
players often fail to make contact with the ball. To ease this
learning process, we added a beginner mode to the game, in
which there is no gravity; the ball moves only in a plane parallel
to the ground. In this mode, the objective of the game is to hit a
target that appears behind the opposing player; this game is more
like Haptic Battle Air Hockey. The reduction in the necessary
movement complexity greatly simplified the game for initial
users. This provided an appropriate starting point for novices.
Similarly, because we found that the spin effect complicates
gameplay, this feature is now optional and modal: the player has
to hit a key to enable the spin paddle, which is represented by a
change in the paddles color.
Another issue addressed by our initial experiments was the
tendency of players to make movements that can potentially
damage the (expensive) haptic devices.
Because HBP is
somewhat immersive and has the general feel of a video game,
players tended to stop treating the device with the normal care
generally used around haptic feedback equipment. The primary
danger was rapid motion through the singularity in the Desktop
Phantoms workspace, which could potentially break the last link
of the device. Therefore, we found it critical to instruct players to
use a particular grip (similar to a traditional table tennis grip)
when playing HBP (see FIGURE 2). This was a bit counterintuitive, since players preferred holding the Phantom as they
would hold a tennis racquet. This inconvenience has potential
implications for the design of haptic devices intended to be used
for video games. Future work should also explore the impact of
device structure and player grip on ergonomics; high-force and
high-frequency applications may have harmful side effects,
particularly for wrist-intensive applications.

5. Results and User Reponses

We have successfully conducted games of Haptic Battle Pong
both locally and across remote sites (including transcontinental
games from Rhode Island to California). The networking strategy
allows seamless transfer of physical simulation control, and thus
smooth haptic interaction at all times.

(a)

(b)

FIGURE 2: The grip used by the player significantly affects the

structural impact that gameplay has on the haptic device.
(a) correct grip, (b) incorrect grip, shown at this devices
singularity

Page 4 of 5

Overall, players felt that HBP provided them with a novel gaming
experience, which required significant skills that had not been
developed through experience with other video games.

6. Conclusion and Future Work

We have developed a game that makes extensive use of the
Phantoms capabilities, both 6-dof input and 3-dof force feedback.
The game can successfully be played across an intranet or
internet, and players are left with a sense of physical interaction
with both the gaming environment and the opposing player.
This success depended on a critical assumption, namely the
private haptic space assumption described in SECTION 4. We
propose that this assumption will generalize well to a broad
variety of video games, and will be critical to allowing massmarket haptic devices to be used successfully over generalpurpose networks. Our first priority in future work in this area is
to apply the HBP simulation environment to other game types, to
support this hypothesis.

MENDOZA, C., AND LAUGIER, C. Simulating Cutting in Surgery

Applications using Haptics and Finite Element Models. IEEE Virtual
Reality 2003, Los Angeles, USA, March 2003.
MORRIS, D., AND JOSHI, N. Alternative Vision: A Haptic and Auditory
Assistive Device. Student poster presented at the ACM SIG-CHI
Conference on Human Factors in Computing Systems, Fort Lauderdale,
USA, 2003.
NAHVI, A., NELSON, D.D., HOLLERBACH, J.M., AND JOHNSON, D.E.
Haptic Manipulation of Virtual Mechanisms from Mechanical CAD
Designs. International Conference on Robotics and Automation,
Leuven, Belgium, May 1998.
RocSearch Ltd. Video Game Industry. Commercial market report.
http://www.rocsearch.com/pdf/Video Game Industry.pdf
SJSTRM, C. Using Haptics in Computer Interfaces for Blind People.
ACM SIG-CHI Conference on Human Factors in Computing Systems,
Seattle, USA, March 2001.
ZHANG, J., PAYANDEH, S., AND DILL, J. Haptic Aided Design: A Case
Study. 11th IEEE Symposium on Haptic Interfaces for Virtual
Environment and Teleoperator Systems, Los Angeles, USA, March
2003.

Additionally, we plan to extend the game to more than two

players; the assumptions we made allow a natural generalization
to an arbitrary number of players.
More sophisticated graphical rendering particularly more
extensive use of stereo rendering may help novice players adjust
to the complex physical input system.
Finally, we plan to incorporate support for other haptic devices in
the near future, particularly the ForceDimension Delta and the
Phantom 6-dof. The latter will allow true torque feedback; we
hope to explore the impact this has on the realism of ball contact.
Additionally, as high-degree-of-freedom haptics gains popularity,
additional work will be necessary to evaluate the ergonomic
impact of high-frequency, complex force-feedback. A future
study will be necessary to evaluate this more formally; this may
impact the ability for haptic devices to break into the consumer
market.
REFERENCES
AGUS, M., GIACHETTI, A., GOBBETTI, E., ZANETTI, G., AND ZORCOLO, A.
Real-time Haptic and Visual Simulation of Bone Dissection. Presence:
Teleoperators and Virtual Environments, 12(1): 110-122, February
2003.
FRITZ, J.P., WAY, T.P., AND BARNER, K.E. Haptic Representation of
Scientific Data for Visually Impaired or Blind Persons. Proceedings of
the Eleventh Annual Technology and Persons with Disabilities
Conference, Los Angeles, USA, April 1996.
GRANGE, S., CONTI, F., ROUILLER, P., HELMER, P., AND BAUR, C. The
Delta Haptic Device. Mecatronics 2001, Besancon, France, July 2001.
MASSIE, T.H. AND SALISBURY, J.K. The PHANTOM Haptic Interface: A
Device for Probing Virtual Objects. Proceedings of the ASME Winter
Annual Meeting, Symposium on Haptic Interfaces for Virtual
Environment and Teleoperator Systems, Chicago, USA, November
1994.

Page 5 of 5