![](bg3.png)
Figure 2: An example of coincidence perception.
3.2 Behavior Construction
Object traces are constructed during the human
computer interaction process and interpreted and
simulated when the operations are required to repeat.
Definition 1: (Temporal behavior segment)
Temporal behavior segment (tbs) is a basic cell of
object behavior, has the form: <t
i
,cn
start
,bhv
k
,
cn
end
>. Let BHV be set of the primitive object
behavior type set
,
CN be a constraint set, T be the
time set. Where t
i
T is the start time of the behavior
segment, bhv
k
BHV is a primitive behavior type,
cn
start
CN is an original constraint, cn
end
CN is the
terminate constraint.
Behavior segment above is used to specify the
flow transitions. Here the typical behavior types
mainly include translation (Trans) and rotation (Rot).
The constraint types mainly include ‘Point’, ‘Line’,
‘Plane’, ‘Sphere’, ‘Circle’, ‘Distance’, and etc.
Definition 2
:(
Short behavior sequence
)
Short
behavior sequence (sbs) is a sequence made up of
three temporal behavior segments which are
respectively corresponding to feature-matching
perception, coincidence constraint perception and
face-mating perception, has the form: tbs
1
;tbs
2
;tbs
3
.
The notation ‘;’ expresses the sequential
operation between two temporal behavior segments.
Short behavior sequence is an abstract of the
assembly operation from the domain. In order to
simplify it, translation motion and rotation motion
are separated apart, that is, when translation is
executed, the degree of rotation freedom is frozen
and vice versa.
Definition 3: (Object trace) Object trace (otrace) is
a successive short behavior sequences from an
initial state to a terminate state, it has the form:
sbs
1
;sbs
2
;…;sbs
n
.
Object behaviour is used to formalize the user’s
operation. In this way, the user’s direct manipulation
is translated into object’s behaviour when an object
is manipulated. The expert’s knowledge is thus
embedded into the object and guides the object to
simulate the human operations in a similar
circumstance. From the object’s assembly trace, we
can easily reverse or repeat the assembly process at
any time.
4 THE EXPERIMENTS AND
ANALYSIS OF 3DIOOM
MODEL
The authors conducted experiments to validate the
provided 3D interaction oriented object model
(3dIOOM). The first aspect is about semantics
construction. IOS represents the metric of system
capability of semantics construction. The second
aspect is about perception performance and behavior
capability. IOB represents the metric of perception
and behavior. The last aspect is interaction load.
IOC represents the metric of total cognition load.
The hardware the authors used is PC machine
and Logitech 3D spacemouse. The software platform
used is Open Inventor 5.0 and Microsoft Visual C++
6.0. A gear case which includes 36 parts is used as
an example for verifying and validating 3dIOOM.
Three models: 3dIOOM, VRML/X3D with AABB
algorithm and VRML/X3D with K-DOP algorithm
are compared because most of the applied models
have the same complexity with them. There were 27
participants. All are regular students at Beijing
Institute of Technology. The participants were
randomly divided into three 9-member groups for
the three experiments respectively. The task is to
assembly 35 parts on the gear case in the virtual
workshop. The metrics are formalized and all values
are normalized in 10 scales. The discrete events and
the evaluated indexes are counted by the programs.
The experiment results are shown by Figure 3.
On the aspect of semantics, 3dIOOM model can
create 8 relationships, i.e., approaching, feature
matching, aggregation, constraint dependency,
coincidence, face mating, and two traces, while
other models only create a collision.
On the aspect of perception and behavior, the
perception efficiency and behavior capabilities on
representation, adaptation are compared. Perceptions
used in 3dIOOM model are real-time. The behavior
mechanism put forward in this paper can make
behavior be constructed directly from participant’s
direct manipulation while others cannot. The
behavior in 3dIOOM is adaptable to the changed
environment, while the behaviors in other models
can not have this capability.
On the aspect of HCI supporting, the cognitive
load in 3dIOOM comes from the sensing of feature
matching searching, coincidence and the face mating.
On the contrary, in models of VRML/X3D
combining with AABB or K-DOP, the load comes
from collision events which occurred a lots of times.
3D INTERACTION ORIENTED OBJECT MODEL
395