AIplanning

AI planning is a branch of artificial intelligence that focuses on automatically generating sequences of actions, or plans, that transform an initial state into a goal state given a model of the world and available actions. A plan consists of actions with preconditions that must hold for the action to be applied and effects that describe how the world changes after the action.

Classical planning studies deterministic, fully observable environments and seeks complete, executable plans. Representations such as STRIPS and the more expressive PDDL describe actions with preconditions and effects, enabling planners to search or infer sequences of actions. Planning approaches include graph-based methods like GraphPlan, search-based planners, hierarchical task network (HTN) planning, and partial-order planning, each with different trade-offs between expressiveness and scalability.

Planning under uncertainty and execution monitoring address environments with nondeterministic outcomes or partial observability. Probabilistic, contingent, and conformant planning extend classical models, and integration with perception, execution monitoring, and reinforcement learning supports replanning when the world diverges from the model.

Applications span robotics, logistics and transportation, manufacturing, space missions, autonomous agents in games, and automated workflow and software orchestration. The field also investigates plan validation, plan repair, and runtime planning to handle dynamic tasks.

History and systems of note include STRIPS (early 1970s), GraphPlan (1990s), and the rise of PDDL-based planners such as Fast Downward. Ongoing challenges involve scaling to large domains, improving plan quality, dealing with uncertainty, and integrating planning with perception and actuation in real time.

The field is commonly referred to as AI planning or automated planning and scheduling (APS).