An explicit assumption during the design of the ATLANTIS architecture was that sensing is error-prone. This facility of error-correction and invocation of alternative methods facilitates the buffering of errors.
Learning
The PCA architecture of Antonisse and Wechsler
supports learning through the modification of schemata.
Supports multiple
simultaneous goals
ATLANTIS supports multiple simultaneous goals through mechanisms of
deliberation which can select a particular goal and plan actions
needed to achieve the goals. During the process of achieving goals, the
primary goal can be changed such as occured in the example experiments.
PCA robots are designed to operate in areas such as "package and mail delivery, hazardous waste cleanup, deep sea and space exploration, security and surveillance, and military reconnaisance tasks." The strict requirement of autonomy in these environments (today) forces extension of the Brooksian subsumption architecture, thereby raising integration issues.
Navigation
The motivation of the control layer is to provide
operators that can perceive and react to the
environment for the purpose of navigation.
Adequacy
<
a>
ATLANTIS agents extend the putative adequacy provided by the
subsumption by adding the feature called cognizant failure.
Coherence
ATLANTIS agents behave coherently in that they are based on subsumption
architecture. The extensions to the architecture enhance the coherence by
providing default contingency actions and the possibility of deliberation.
Saliency
The features of ATLANTIS seem appropriate for the environment but some the future mechanisms may not be
needed. All of the robotic features seem useful.