Académique Documents
Professionnel Documents
Culture Documents
*Adaptive Machine Technologies Inc., 1218 Kinnear Road, Columbus, Ohio, USA
** 1078 Trowbridge Court, Longwood, FL 32750, USA
107
concepts as beliefs, desires, and goals. Agents derive from the overall capabilities of the
believe propositions about the world. Agents have agent (e.g., a robot firefighter), the
desires about the world. When an agent decides to characteristics of the environment (e.g., a fire
actively pursue a desire, it forms an appropriate site), and the requirements of the task (e.g.,
goal. In this paper, we will equate desires with putting out the fire).
goals for simplicity. We assume that if an agent
desires something, it immediately forms a The flexibility of these time constraints (hard
corresponding goal. In Shoham 's language, for deadlines vs soft deadlines), and the degree to
example, which they can be traded for various levels of
quality of the result, depend on the characteristics
< 9: 15, BHarold < 9:30, GC.l'io < 10:00, of the real world from which they are derived. For
weld( Calvin, Car14 ) > > > example, it is difficult to dynamically change the
natural frequency of a physical robot arm, so the
means that at 9:15, Harold believes that at 9:30, arm's servo controller must operate near a given
Calvin will have the goal of welding Car14 at 10:00. frequency. Fortunately, at the servo level, the arm ' s
When parsed in an LR fashion, such a statement world is fairly static and well understood. The servo
can be viewed as being a hierarchy of beliefs and computations usually take a fixed or tightly bounded
goals. However, at any point during the execution amount of time . In contrast, the task planner for a
of the application, there may exist a discrepancy firefighting robot is operating in a more dynamic
between the beliefs (goals) of one layer and the and less understood world. It is not possible to plan
goals (beliefs) of an immediately neighboring layer. optimally, using any fixed amount of computing
Such discrepancies result from race-conditions, resources, since there is always too much
timing-skews, or simply delays in propagating information, some of which is wrong, and some of
information through the hierarchy. Therefore, when which will become outdated by new information.
such high-level statements are translated into lower- Fortunately, the task planner may be able to
level constructs, they must be augmented with generate acceptable sub-optimal plans, as time and
timing constructs that capture the acceptable resou rces allow.
granularity of belief (goal) discrepancy.
In real-time applications, time constraints are The Valid Time is the time at which the State is
derived from the entities in the real world (e.g., supposed to correspond to the state of the
agents, environments and tasks) and the associated real-world entity . The Generation
relationships among them . In robotics, for example: Time is the time at which the tuple was created.
This creates a two-dimensional space of values for
Servo-control time constraints derive from the each attribute . The States themselves may be
structures and dynamics (e.g., natural complex entities, containing, for example, the
frequencies) of the objects being controlled expected value, tolerances, probabilities, and so on .
(e.g., the aluminum links of a robot
manipulator), and how they interact. Objects are typically composed of sub-objects.
Figure 2 shows the sub-objects of the Arm in Fig. 1.
Elemental-move motion-planning (i.e ., The attributes of such a composite object may be
planning how to move from point A to point synthesized from those of its sub-objects (e.g.,
B) time constraints derive from the position mass), or they may reflect features unique to the
and velocity required of the object and the composite object which are not found in the sub-
positions and velocities exhibited by other objects. For example, an Arm can "lift" a payload,
objects (obstacles). but a Gripper cannot.
108
A rm-P osition to the Planner ' s Believed -Arm-
PLANNER Desired
ARM Position a nd transmission of an Inform message
De sired
A rm
Positio n
~
.. A rm
Positi on
containing t he value of the Pl anner's Desired-Arm-
Position to the Arm ' s Desired-Arm-Position.
Believe d
Arm
...
r-.
Be li eved
A rm AN EXAMPLE
Positi on Pos ition
In this sectio n, we de scribe an example application,
De sire d
Grippe r
Positi on
.. Desired
--"" G rippe r
Position
s pecify its required behavior, a na lyze the
information content and flow, a nd di sc uss de s ign
and implementation alternatives.
Beli eved
......
Beli eved
Grippe r Gripper
Pos iti on Pos iti on Task Specification
JOINT CONl ROLLER Figu re 3 s hows a po ss ihle sequence of events for the
task. Based on its beliefs at time T(O). the Pl a nner
I----.t Joinl decide s to begin the two-step p lan . At T(l). the
Posilion
Planner inform s the Arm th at it de s ire s th e A r m to
be in pos iti on P(drop) at t im e T(drop). At T(2).
the Arm begin s the ac ti v itie s necessa ry to move the
Arm. Thi s may inc lud e iterati o ns of the Arm 's own
Choose-Execute-Eva lu ate cycle. made up of s m a ll-
GR IPPER grai n motion p lan ning , sens in g , and servo control.
The Arm actually fini s hes movin g to P(drop) at
T(move done). Thi s time may va ry. d epe ndin g on
th e previou s pos ition of the Arm.
109
where T( 4) is the Generation Time, T( drop) is the
Valid Time, and P(drop) is the State .
PLANNER ARM
T(O)
The transfer of desires and beliefs between objects
via Inform messages consists of transferring
Decide 10 Move Arm
attribute values. For example, the statement :
T(J)
"At T( 1), the Planner informs the Arm that it
T(2) desires the Arm Position to be P(drop) at
T(drop) . The Arm receives the information
T(3)
at T(2) ."
Notice that only the < Valid Time, Value > sub-
tuple is transferred . This amounts to the Arm
"internalizing" the desires of the Planner - it
Fig. 3. An interaction sequence. accepts the Planner ' s desires as a c ommand. That
is, the Arm records the statement:
T(move done) varies, it may be necessary to trade "At T(2), the Arm desires the Arm Position
the timeliness of the task (i.e., opening the gripper to be P(drop) at T(drop) ."
at exactly T(drop for the accuracy of the Planner's
beliefs about the Arm ' s position. Several options In general, it is possible to have arbitrarily complex
exist: attributes, such as the Arm attribute:
110
T(drop), there are two options: the functionality (and belief inconsistency
problems) of the Planner into the Arm. In fact, the
The Arm could send "Put(Planner, original Planner-Arm system can be thought of as
Believed-Arm-Position, T( drop ),Position)" to encapsulated in a larger "Smart-Arm" object.
the Planner each time the Arm's belief about
its position at T(drop) changes . The Arm
controls when the information is transferred. CONCLUSION
The Planner could periodically send Intelligent, distributed, real-time systems can never
"Get(A rm, Be lieved-Arm- Posi tion, T( drop " maintain perfect, consistent information about a
to the Arm . The Planner controls when the complex and changing environment, particularly
information is transferred. given the hard time constraints on their operation .
System specification and design languages which
The first option may make sense when the Arm ' s represent the meanings of timeliness and quality
helief changes significantly at well-defined points . and which can be mapped to hard-real-time
When asked to move to P(drop) by T(drop), for implementations are needed.
example, the Arm might first determine if P(drop)
is in its possible range of motion; then determine if We have proposed a particular object-oriented
the move from its current position to P(drop) is system model that can represent high-level concepts
possible by T(drop) ; then execute the move. The such as beli efs and desires and have discussed the
Arm's belief may become more certain after each translation of these concepts to practical designs
step, and pass the updated information to the and implementations .
Planner.
Dynamically trading belief and desire consistency
The second option may make sense when the Arm for timeliness may allow tasks which cannot be
itself cannot judge the significance of its beliefs . performed on-time and with complete confidence in
This is true, for example, when the Planner is the results to be carried out slightly late or with less
gathering general information about the Arm's state confidence. Practical approaches to this problem
before it puts any plan into action. are needed .
III