AI is a critical technology for the future. Even if it's intelligent robots, self-driving cars, or smart cities, they will all use various aspects of AI! However, to create any sort of AI project, careful planning is required. 

What is a Plan?

We require the domain description, action specification, and goal description for any planning system. A plan is assumed to be a series of actions, each of which has its own set of preconditions that must be met before acting, as well as some unintended consequences that can be either positive or negative.

Why is it Important?

• Automation is a rapidly growing trend that necessitates effective automated planning.
• Numerous industrial applications of planning (e.g. robots and autonomous systems, cognitive assistants, cyber security, service composition)

Planning in AI

Planning is the process of determining a procedural path for a declaratively described system to follow to accomplish its objectives while optimizing overall performance measures. Automated planners select the transformations to apply in a given state from a set of possible transformations. In comparison to the classification problem, planners guarantee the quality of the solution. Planned actions are thus thought of as the reasoning side of action. In other words, planning is all about determining the actions to be taken by the AI system and the system's functioning in domain-independent situations.

Thus, at the most fundamental level, we have Forward State Space Planning (FSSP) and Backward State Space Planning (BSSP).

1. Forward State Space Planning (FSSP)

FSSP operates similarly to forward state-space search. It states that given a start state S in any domain, we take the necessary actions to acquire a new state S' (which includes some new conditions), which is referred to as progress, and this process continues until we reach the goal state. The actions must be appropriate in this instance.

• Con: Excessive branching
• Pro: An advantage is that the algorithm is sound.

2. Backward State Space Planning (BSSP)

BSSP operates similarly to backward state-space search. This progresses us from the goal state g to the sub-goal g', which is determining the previous action required to accomplish that goal. This is referred to as regression (moving back to the previous goal or sub-goal). These sub-goals must also be examined for consistency. In this instance, the actions must be pertinent.

• Con: The disadvantage is that the algorithm is not sound (sometimes inconsistency can be found)
• Pro: The advantage is that the branching factor is small (very small compared to FSSP)

Thus, to create an efficient planning system, we must combine the characteristics of FSSP and BSSP, which results in Goal Stack planning.

3. Goal Stack planning

This is one of the most critical planning algorithms, and it is used exclusively by STRIPS.

In an algorithm, the stack is used to store the action and ensure that the goal is met. A knowledge base is used to store information about the current state and actions.

A goal stack is analogous to a node in a search tree, where branches are created when an action is selected.

Conclusion:

The purpose of this article is to educate the reader about the planning and learning concepts used in AI. The capacity to generate knowledge and to apply that knowledge in a practical context is a critical characteristic of any intelligent system. Automated Planning and Learning is a research paradigm that focuses on the development of intelligent systems and technologies that combine the capacity to make decisions. Moreover, planning in AI is to generate courses of action (i.e., plans) about prior experiences and future problems that the system must address.