Object Associations A Simple and Practical Approach to Virtual 3D Manipulation Richard W. Bukowski Carlo H. Skquin University of California at Berkeley 1 Abstract This paper describes a software framework to aid in design- ing and implementing convenient manipulation behaviors for objects in a 3D virtual environment. A combination of al- most realistic-looking pseudo-physical behavior and ideal- ized goal-oriented properties, called object associations, is used to disambiguate the mapping of the 2D cursor motion on the display screen into an appropriate object motion in the 3D virtual world and to determine a valid and desirable final location for the objects to be placed. Objects selected for relocation actively look for nearby objects to associate and align themselves with; an automated implicit grouping mechanism also falls out from this process. Concept, struc- ture, and our implementation of such an object association framework in the context of the Berkeley Soda Hall WALK- THRU environment are presented. 1 Introduction Creating a fully equipped model of a large, furnished build- ing for virtual walkthroughs is an arduous task. Even as- suming the availability of a good interactive 3D geometry editor with a friendly and efficient user interface, such tasks are inherently much more difficult than drafting and edit- ing in only two dimensions. The problem with a 3D world is that it is impossible to exactly control all six degrees of freedom (DOF) at once with only 2-dimensional input and display devices. Typically, software solutions are used to map 2D cursor motion to limited 3D object space motion [12]. These can be cumbersome to use in complex environ- ments, and do not address the fact that objects often require positioning with respect to objects around them. High-tech solutions such as the “SpaceBall” [2], “DataGlove,” 3D mice [15], or virtual 3D displays do not solve the problem either; precise placement of objects in three dimensions is hard - even in the real world - unless we get help from the phys- ical interactions of the objects we want to place. Consider positioning a picture frame one millimeter in front of a wall ‘Computer Science Division, Soda Hall, Berkeley, CA 947?0-1776; bukowskiOcs.berkeley.edu and sequinOcs.berkeley.edu. Permission to copy without fee all or part of this material is granted provided that the copies are not made or drstnbuted for direct commercial advantage, the ACM copyright notrce and the title of the publication and Its date appear,. and notice IS given that copying is by permission of the Assocration of Computing Machinery. To copy otherwise, or to republish, requires a fee and/or specific permission. 1995 Symposium on Interactive 30 Graphics, Monterey CA USA 0 1995 ACM O-89791 -736-7/95/0004...$3.50 without touching the wall with the frame or with your hands; visual feedback alone cannot do a satisfactory job. As part of the Berkeley WALKTHRU Project we have built a prototype version of an object manipulation system, called “WALKEDIT,” tailored to populating large building models with furniture, personal items, books, coffee cups, and various trimmings and details that make such a build- ing model look real and interesting (see Figure Cl). Our approach is based on a system of “object associations,” a software framework that supports simple and practical ma- nipulation of 3D objects with 2D I/O devices via two spe- cial types of programmer-supplied procedures and an im- plicit grouping behavior. It gives the programmer the abil- ity to specify object-dependent methods of. disambiguating 2D gestures in a 3D world and allows association of suitable local behavior with database objects to make precise default placement easy. These associations usually fall somewhere between physical simulations and mathematical constraints, but can be less formal and more flexible than either. 2 Interactive Building Environments In the process of developing an editor for our Soda Hall WALKTHRU program, we examined different methods for helping the user to move objects in 3D with 2D devices. We wanted the process of moving furniture in a 3D virtual environment to be as quick and easy as moving cut-out card- board pieces on a floorplan. However, it should also be pos- sible to force objects to align themselves nicely to walls and to one another, if the operator chooses such an option. There is no single correct answer to the question of what “ideally” should happen in response to a user dragging an object across the 2D display screen. There is at least one uncontrolled degree of freedom (DOF) due to the third co- ordinate of the virtual world. Choosing the “right” value to be assigned to this coordinate becomes a contention between realistic (physically correct) and teleological (.goal-oriented) models for the virtual world, and is strongly dependent on the specific application domain. Traditional tools tend to take an extreme stand on one or the other end of the spec- trum. Most 2D drafting tools provide an idealized goal-oriented behavior. Selected shapes freely follow the cursor “across” other objects and snap nicely into alignment with other fea- tures if grids or gravity have been turned on. Setting up these extra controls requires some overhead for activating align- ment manifolds, setting up tugboats and orientation frames, changing editing modes, or grouping and un-grouping ob- jects. In 3D virtual worlds, the situation is even worse; ob- 131jects now have twice the number of DOFs to be controlled. However, the real world being modeled can often provide disambiguating clues or implicitly desirable alignments. By exploiting these application-specific expectations, some of the