Page 281 - Introduction to AI Robotics

264
7
The Hybrid Deliberative/Reactive Paradigm
mans to use alternative sensors and effectors (e.g., opening a door with
the other hand when the holding something in the preferred hand).
CARTOGRAPHER A Cartographer which is responsible for creating, storing, and maintain-
ing map or spatial information, plus methods for accessing the data. The
cartographer often contains a global world model and knowledge repre-
sentation, even if it is not a map.

MISSION PLANNER A Mission Planner which interacts with the human, operationalizes the
commands into robot terms, and constructs a mission plan. For example,
the ideal robot assistant might be given a command: “Lassie, go get the
sheriff.” The mission planner would interpret that command to mean to
first physically search for a person, identify him as a sheriff because of
the uniform, attract his attention, and then lead him back to the current
location. The mission planner might have access to information that sher-
iffs are most likely to be in their office, donut shops, or the last place they
were seen. The plan might first be to navigate to the sheriff’s office.

PERFORMANCE A Performance Monitoring and Problem Solving agent which allows the robot
MONITORING AND to notice if it is making progress or not. Notice that this requires the robot
PROBLEM SOLVING
to exhibit some type of self-awareness.

7.3.2 Styles of hybrid architectures

MANAGERIAL STYLES Architectural styles can be loosely divided into three categories. Managerial
styles focus on subdividing the deliberative portion into layers based on the
scope of control, or managerial responsibility, of each deliberative function.
A Mission Planning module would be able to direct other, subordinate de-
liberative modules such as navigation, because Mission Planning (where to
STATE HIERARCHIES go) is more abstract than Path Planning (how to get there). State hierarchies
use the knowledge of the robot’s state to distinguish between reactive and
deliberative activities. Reactive behaviors are viewed as having no state, no
self-awareness, and function only in the Present. Deliberative functions can
be divided into those that require knowledge about the robot’s Past state
(where it is in a sequence of commands) and about the Future (mission and
MODEL-ORIENTED path planning). Model-oriented styles are more nebulous. They are character-
STYLES ized by behaviors that have access to portions of a world model, often to the
point that they appear to have returned to the Hierarchical monolithic global
world model.