Summary

Summary

As defined in Ch. 4, a reactive implementation consists of one or more behaviors,
and a mechanism for combining the output of concurrent behaviors.
While an architectural style (subsumption, potential fields) may specify the
structure of the implementation, the designer must invest significant effort
into developing individual behaviors and into assembling them into a sequence
or an abstract behavior.
Schema theory is highly compatible with Object Oriented Programming. A
behavior is derived fromthe schema class; it is a schema that uses at least one
perceptual andmotor schema. If a behavior is composed of multiple schema,
it must have a coordinated control program to coordinate them. Finite State
Automata offer a formal representation of the coordination logic needed to
control a sequence of behaviors. Scripts are an equivalent mechanism with a
more natural story-like flow of control.
The steps in designing robot intelligence under the Reactive Paradigm are:
1. Describe the task,
2. Describe the robot,
3. Describe the environment,
4. Describe how the robot should act in response to its environment,

5. Refine each behavior,
6. Test each behavior independently,
7. Test with other behaviors,
and repeat the process as needed. A behavior table serves as a means of representing
the component schemas and functions of the behavioral system. For
each behavior, it shows the releasers, the motor schemas, the percept, and
the perceptual schemas.
These steps emphasize the need to fully specify the ecological niche of the
robot in order to design useful behaviors. Since the idea of behaviors in the
Reactive Paradigm is derived from biology, it is not strange that the idea of
a robot being evolved to fit its environment should also be part and parcel
of the design process. Regardless of the implementation of the coordination
program, the control should rely on the world to inform the robot as to what
to do next, rather than rely on the program to remember and maintain state.