Let's play pool on the Responsive Workbench! (tm)

		Robert Laddish, Cyprien Godard, Alan Steremberg


Introduction:

   Our project is to produce a 3D virtual pool table on the responsive
   workbench. We are planning on using a physical cue which extends to a
   virtual cue over the table. The playing field will be above the
   surface of the workbench and inset from the sides to extend the 3D
   illusion to the entire virtual pool table. This will give room to
   display 3D imagery from strange angles and display the virtual cue.
   
   This project proposal has been broken down into the following sections:

   Team
   Functional Breakdown
     1. Physics:  [ Cyprien ]
     2. Visuals: [ Rob ]
     3. Game dynamics [ ??? ]
     4. Tracking: head and cue. [ Alan ]
   Progress
   Time-line

Team:

   Our development team consists of the following people:
  
   Alan Steremberg alans@xenon
   Cyprien Godard  cyp@cs
   Robert Laddish  robl@leland

Functional Breakdown:

   We've broken our project down into 4 functional areas: physics, visuals,
   game dynamics, and tracking.

1. Physics:  [ Cyprien ]

   We would like to model each ball as it's own object, and model the
   interaction between them separately. For each ball, we will have to
   take into account mass, velocity, center of inertia, acceleration,
   point of contact to the ground. When two balls collide we will have to
   compute the new directions and velocities of the balls after
   collision. We will have to find the position at which the cue hit the
   ball and find a paradigm for the contact cue-ball. We will also have
   to model friction so the balls do not continue to roll forever.
   
   We will implement a 3D collision detection package between the balls
   and between the balls and the bumpers and between the cue and the
   bumpers and the balls.
   
   Initially we might do a simple 2D physics package, and not allow the cue
   to cause the cue ball to jump. If this runs quickly enough, we will
   attempt to model other 3D parameters, and allow the cue-ball to jump off
   of the table. Marc suggested that we find a 3D physics package, but
   Berndt thought that we wouldn't be able to find one that ran in real time
   and that we would have to implement it ourselves. We will check with
   some key people from the Robotics lab first, and if no package can be
   found, we will implement our own.

2. Visuals: [ Rob ]

   We would like to be able to display the pool table as well as the
   moving balls in 3D using the workbench libraries. For the mock demo,
   each ball will have a different color. After this is achieved,
   we will add texture (color + numbers) and shadows. Likewise, the virtual
   cue, the corners of the table, and the pockets will be textured and
   shadowed as well.

   The pool table will have to be of the correct aspect ratio, and the balls,
   cue, sides, and pockets will likewise have correct dimensions. As noted
   in the game dynamics section, we will allow the pool table to be rotated.

   The virtual table will have to be inset from the RWB side to allow side
   features to retain their 3D appearance from most angles. It will also
   allow us to extend the physical cue as a virtual one. Berndt claimed that
   the 3D effect would be better above the table, and this will help prevent
   scratching the tables surface with the physical cue. Marc claimed that
   it would be better to render the table below the surface. It is worth
   noting that some pool players place the cue between the knuckles of one
   of their hands for stability. This will be very difficult to do if the
   virtual surface is above the RWB table, and impossible if the virtual cue
   needs to be placed between the knuckles. We plan to very the height and
   the length of the virtual cue to achieve optimal playability.

   We expect problems due to the graphics redraw rate of fast moving balls.
   If the physical cue is allowed to obstruct the RWB, there may be 
   undesirable side effect.

3. Game dynamics [ ??? ]

   The game dynamics will take care of setting up the board, telling players
   to alternate, detecting winning conditions, and presenting different games
   of pool to be played to the users. It will detect "scratch" balls and allow
   the user to reposition them on the table. For the mockup, we will not demo
   any particular game, and allow users to take balls from pockets and place
   them anywhere on the board they wish.

   Because the transmitter is at a fixed location, we will also need to allow
   the user to rotate the table. This may be limited to 180 degree rotations
   only to maximize the virtual table size.

   We may allow users to save & reload games as well.

4. Tracking: head and cue. [ Alan ]

   We will be tracking the users head with the crystal eyes glasses,
   and we will track the end of the physical cue with a polhemus
   sensor. Initially, we will use the sensor without the physical cue.

   The main problem we expect to run into is lag of reading the
   sensors and updating the display. This may not be much of a problem
   when orienting the virtual cue, but hitting the cue ball is a fast
   action motion. One possible workaround is to implement a virtual
   slider used to select the force of the cue when it hits the ball. Our
   original idea was to perform numerical differentiation to achieve a
   guess at acceleration to compute the force, but this may not be
   possible.

   We also expect some placement errors between the physical and virtual
   cues.