How does a robot figure out the steps to move a box from one room to another? Or how does a logistics system create a plan to get a package from a warehouse to your door? The answer lies in AI Planning, a field that was kicked off by a system called STRIPS. It was originally designed to control "Shakey the Robot," one of the first robots to reason about its own actions.
🔍 The Discovery
Name of the Technology: STRIPS (Stanford Research Institute Problem Solver)
Original Creator/Institution: Richard Fikes and Nils Nilsson at the Stanford Research Institute (SRI).
Year of Origin: 1971
License: The formalism is a public domain concept.
The genius of STRIPS is its simple way of describing the world. An "action" is defined by three things: its preconditions (what must be true to perform the action, e.g., "the robot must be in the same room as the box"), its add list (what becomes true after the action, e.g., "the robot is now holding the box"), and its delete list (what is no longer true, e.g., "the robot's hand is now empty"). By providing an AI with a list of these actions, a starting state, and a goal state, a "planner" algorithm can automatically search for and find a valid sequence of actions to achieve the goal. It's a recipe for automated problem-solving.
🛠️ Ready for Today: Why This Isn't Just Theory
The STRIPS formalism became the foundation for the standardized Planning Domain Definition Language (PDDL), which is used in an annual International Planning Competition. This has resulted in decades of research and development, producing highly optimized, open-source planners that can solve incredibly complex problems in seconds.
Status: The concepts are public domain, and the PDDL standard is open.
Implementations: You don't need to write a planner; you just need to describe your problem and feed it to a state-of-the-art solver:
Fast Downward: A popular and powerful open-source planner that is a frequent competition winner. You provide it with PDDL files describing your domain and problem, and it outputs the solution plan.
Online Editor & Solver: The website
planning.domainsprovides an online editor where you can write PDDL and solve it with various planners directly in your browser.Python Bindings: Libraries like
pyperplanallow you to call planners from within a Python environment, making it easier to integrate into a larger application.
💡 Creative Applications (Ideas To Get You Thinking)
AI Planning is used heavily in robotics and logistics, but those are complex, crowded fields. The power of a general-purpose planner is that it can solve any problem that can be described with actions and states, opening up many non-obvious, low-competition opportunities.
Idea 1 (Automated Tutorial & Onboarding Generator): Build a tool for software developers or UX designers that automatically generates step-by-step tutorials. The designer defines all the possible user actions in an app (the "domain") and a goal (e.g., "user has created a new document"). The planner then automatically finds the shortest, most logical sequence of steps to get the user there, generating a perfect onboarding guide.
Idea 2 (Dynamic "Recipe" Generator for Complex Workflows): Create a tool for fields with complex, multi-step processes like cooking, DIY projects, or scientific experiments. A user selects a goal ("I want to bake a sourdough loaf") and their starting conditions ("I have flour, water, but no active starter"). The system, using a pre-defined knowledge base of actions, generates a custom plan, even including steps like "create a starter," which wouldn't be in a static recipe.
Idea 3 (Narrative and Quest Generation for Games): Develop a plugin for game engines that uses a planner to generate dynamic quests. A game designer defines the state of the world and a set of possible character actions (e.g., "talk to NPC," "get item," "unlock door"). The system can then generate emergent, logical quest lines to move the story from state A to state B, creating far more replayability than hand-scripted quests.
🐰 The Rabbit Hole
Dive Deeper: The "GeeksforGeeks" website has a clear and well-structured article on the STRIPS planning algorithm. It breaks down the core components—states, goals, and actions with their preconditions and effects—and walks through the classic "blocks world" problem to show how a plan is generated.
Join The Search
Our mission is to unearth the world's most powerful, overlooked ideas. If you know of a technology that is trapped in a niche, overshadowed by hype, or simply deserves a bigger spotlight, please submit it for a future issue here.
Till next time,