Context-based Reasoning (CxBR) is a paradigm intended to model human tactical
behaviors. CxBR is based on the assumption that humans think in
terms of contexts. Contexts encapsulate knowledge about appropriate actions
needed to address specific situations. The CxBR paradigm is composed of a
tactical agent, mission context, major contexts, sub-contexts and sentinel
rules which control the transitions between contexts.
CxBRAgent was implemented using eight different context constructs: 1)
The ExploreContext is the default context of the mission. 2) The
BackTrackContext is called from ExploreContext when there is nothing new to
explore in the map at the current location of the agent. The agent will then
retrace its step and search for new places to explore. 3) The AttackContext is
deployed when there is a hostile entity within the sensor range. 4) The
AvoidContext represents the case when there is a hostile entity within the
sensor range and the agent cannot attack the other agent. The agent will move
away from the hostile agent, trying to avoid being chased or attacked by the
other agent.
|
|
5) The EatContext is called from either the ExploreContext or
BackTrackContext when there is food within the sensor range. The agent will
move towards the food and invoke the eat command on the food resource. 6) The
FleeContext is invoked when the agent have been attacked. The agent attempts to
flee away from an attacker. 7) The MateContext is invoked rules when the agent
can mate. 8) The NearDeathContext is invoked when the energy level of the agent
is below a certain threshold. It will attempt to extend it’s lifeline by
spawning another agent.
The CxBR agent implements a simple path-planning algorithm which allows it to
navigate an internal representation of the global FFM map. The same
path-planner is used when approaching objects in the agent's local sensor
range.
|
