planninglanguage

Planning language is a formal notation used to describe planning domains and planning problems for automated planning and scheduling. It provides a vocabulary of predicates, types, and objects along with action schemas that define how the world can change. A planner uses this model to search a state space for a sequence of actions that achieves the described goal.

States are usually described by ground or parameterized predicates. Actions (or operators) specify preconditions that must hold for the action to apply and effects that change the state. Some planning languages also support numeric variables, duration, resources, or temporal constraints.

Model structure typically splits modeling into a domain description and a problem description. The domain defines predicates, types, and action schemas; the problem provides a set of objects, the initial state, and the goal condition. Many planners require grounding or instantiation of actions for specific problems, though some support lifted representations.

One of the earliest influential planning languages is STRIPS. The current de facto standard for classical planning is the Planning Domain Definition Language (PDDL), with extensions for typing, numeric fluents, durative actions, temporal planning, and conditional or disjunctive goals. Other families include SAS+ and HTN-based languages; probabilistic planning languages extend PDDL with stochastic effects.

In practice, planning languages are used in AI planning research, robotics, logistics, manufacturing, and game AI. The domain author writes the domain and problem files, a planner processes them, and a generated plan is executed or validated. Modeling choices affect solvability and planner compatibility.