ABSTRIPS Developed by Earl Sacerdoti at SRI International
In 1974, Earl Sacerdoti at SRI International published ABSTRIPS, an extension of the STRIPS planning system that organised problem-solving into a hierarchy of abstraction spaces, allowing a planner to resolve high-level constraints before committing to fine-grained detail.
Background
By the early 1970s, automated planning had a working foundation in STRIPS, a system built at SRI International that could reason about sequences of actions to reach a goal. STRIPS represented the world as a set of logical conditions and described each action by what it required and what it changed. Given a starting state and a target state, the planner would search for a sequence of actions that got from one to the other. It worked well enough on small problems.
The trouble was search. As problems grew more complex, the number of possible action sequences exploded. A planner had no good way to decide which details mattered early and which could be left until later. It would try to satisfy every precondition of every action at the same level of effort, whether that precondition was a fundamental constraint or a minor housekeeping step. The result was that planners wasted enormous amounts of work resolving trivial details before they had even established whether the broad shape of the plan was sound.
No obvious mechanism existed to tell a planner: settle the hard things first, and fill in the rest later. That gap is what Sacerdoti set out to close.
What happened
In 1974, Earl Sacerdoti at SRI International published “Planning in a Hierarchy of Abstraction Spaces” in the journal Artificial Intelligence (volume 5, number 2). The system he described, ABSTRIPS, short for ABSTRact STRIPS, added one key idea to the STRIPS framework: a criticality value attached to each precondition of each action.
The criticality value was a number that reflected how fundamental a precondition was. A high value meant the condition was difficult to achieve and central to whether a plan could work at all. A low value meant it was a peripheral detail. By ranking every precondition this way, Sacerdoti could slice the planning problem into a series of abstraction spaces. At the top level, ABSTRIPS considered only the most critical preconditions and sketched a coarse plan that satisfied them. Then it moved down a level, reintroduced the next tier of preconditions, and refined the plan. It continued until the lowest abstraction space, where all preconditions were in play and every detail had been resolved.
The practical effect was that the planner committed to the broad structure of a solution before spending effort on its fine grain. If a high-level plan turned out to be unworkable, the system discovered this early and could backtrack before it had invested time in lower-level detail that would be thrown away anyway. Nilsson, writing about this period in The Quest for Artificial Intelligence, describes the approach as a genuine advance in how planners could organise their search. ABSTRIPS remained a research system rather than a deployed application, but its design shaped how the field thought about structuring search through complex action spaces.
Why it mattered
ABSTRIPS introduced hierarchical abstraction as a practical mechanism in automated planning, demonstrating that decomposing a problem by level of detail could dramatically reduce the search space a planner must explore. This principle (solving abstract versions of a problem first, then refining) became foundational in AI planning research and influenced later work on hierarchical task networks (HTNs). The system also advanced understanding of how knowledge about action preconditions could be structured to guide search more efficiently.
People
Organisations
Sources
- Planning in a Hierarchy of Abstraction Spaces.Artificial Intelligence (Elsevier), Vol. 5, No. 2.Primary source
- The Quest for Artificial Intelligence.Cambridge University Press / Nils J. Nilsson.Secondary
Cite this page
AI Achievements. (1974). ABSTRIPS Developed by Earl Sacerdoti at SRI International. Retrieved 2026-08-22, from https://achievements.ai/milestone/abstrips-developed-by-earl-sacerdoti
@misc{achievements_abstrips_developed_by_earl_sacerdoti,
title = {ABSTRIPS Developed by Earl Sacerdoti at SRI International},
author = {{AI Achievements}},
year = {1974},
url = {https://achievements.ai/milestone/abstrips-developed-by-earl-sacerdoti}
}