Robotics Reference
In-Depth Information
must also be possible to remove goals that cannot be met from the
planning process, along with any partial plans and constraints they
required.
6. The robot has long-term motivations, which enable it to generate
goals, prioritise them, and select the best plan to achieve them.
The key to the overall architecture shown in Figure 44
13
is the robot's
motivations, which can be thought of, depending on the problem do-
main, as its long-term aims or objectives, or as its drives or emotional
states. For example, a robot might be motivated by hunger (the need to
recharge its batteries), by tiredness (the need to rest in order to cool down
its motors), or by curiosity (the need to discover knowledge that might
be relevant to a problem solving task).
A robot's motivations also have another purpose, allowing the robot
to evaluate the plans it generates in order to achieve its goals, and to
choose between alternative plans. For example, a plan to acquire food
that involves a robot walking to the shops to buy a pizza, might con-
flict with its tiredness (lack of battery power) and cause it to explore an
alternative plan that involves cooking something it already has at home.
An important question here is, “when exactly should a robot's mo-
tivations be updated and new goals generated?” In Aylett's design this
happens only after an action has been executed, but in certain types of
situations, and robot soccer is a good example, things happen so quickly
that it is necessary to update a robot's motivations even before one of its
actions has been completed. In soccer, when a player has decided to pass
the ball to a particular teammate, immediately after he has kicked the
ball the passing player will plan what he is going to do, and it is only
by such anticipatory action that a player can maximize his usefulness on
the soccer field. Thus a robot's actions can affect its motivations. The
specification of an action also determines which robot (or robots) is (or
are) intended to execute it.
The flow of the planning algorithm can be followed by reference to
Figure 44. A robot first chooses the best partial plan at its disposal. When
deciding whether to plan to satisfy a goal or to execute an action, the
robot first selects the “best” goal from those that it needs to achieve, and
then determines which actions are ready to be executed.
In selecting
13
Figure 1 of R. S Aylett, A. M. Coddington, and G. J. Petley. “Agent-Based Continuous Plan-
ning”
19th Workshop of the UK Planning and Scheduling Special Interest Group (PLANSIG 2000)
(http://mcs.open.ac.uk/plansig2000/Papers.htm).
