Knowledge Items (KI)

The main KIs that we have used in PELEA are:

• stateL: low-level state composed of the sensory information.

• stateH: high-level state, that gets translated from stateL as an aggregation or a generalization of low level information.

• goals: the set of high-level goals to be achieved by the architecture.

• metrics: the metrics that will be used in the high-level planning process.

• planH: set of high level plans. Each high level plan is a set of ordered actions resulting from the high-level planning process. Usually, they will be sequentially ordered, though parallel plans can also be given.The actions of these plans can also be the goals for the low-level planner (in case we want the low-level planner to act as a dynamic translation mechanism for high-level actions).

• planL: set of low level plans. Each low level plan is again an ordered set of actions resulting from the low-level planning process. Usually, it will consist of only one plan, and several actions that can be executed in parallel. These actions should be operational: directly executable in the environment.

• domainH: definition of actions for high-level planning.

• domainL: definition of behaviors (skills) for low-level planning.

• learning examples: to be used by the learning component to acquire knowledge for future planning episodes, either in the form of heuristics, domain models, or knowledge on the problem specification.

• heuristics: in different forms (control rules, policies, cases, macro-actions, etc.) allow the planners to improve their efficiency in solving future planning episodes.

• monitoringInfo: meta knowledge on the plan that helps to perform the monitoring (as, for instance, the footprint of each action)

Back: Where to use PELEA ?