Davis–Putnam–Logemann–Loveland algorithm

Davis–Putnam–Logemann–Loveland algorithm
Tamkin04iut · CC BY-SA 3.0 · source
Name	Davis–Putnam–Logemann–Loveland algorithm
Author	Martin Davis; Hilary Putnam; George Logemann; Donald Loveland
Introduced	1960s
Application	Automated theorem proving; Boolean satisfiability
Input	Propositional formula in conjunctive normal form
Output	Satisfiable assignment or unsatisfiable

Davis–Putnam–Logemann–Loveland algorithm is a backtracking search procedure for deciding the satisfiability of propositional logic formulas in conjunctive normal form. It originated in the context of automated theorem proving and influenced the development of modern SAT solvers used in verification, artificial intelligence, and electronic design automation. The procedure combines systematic variable selection, unit propagation, and backtracking to determine a satisfying assignment or a certificate of unsatisfiability.

Introduction

The procedure was introduced by authors associated with research traditions at Princeton University, Harvard University, Massachusetts Institute of Technology, and Bell Labs in the mid twentieth century. It built on earlier efforts in symbolic logic linked to figures such as Alonzo Church, Emil Post, Kurt Gödel, and Alan Turing, and it anticipated algorithmic techniques later exploited by projects at Stanford University, Carnegie Mellon University, University of California, Berkeley, and IBM. The algorithm forms a conceptual bridge between classical proof procedures and modern decision procedures employed at institutions like Microsoft Research and Google.

History and development

The original work emerged during an era of active research at institutions including Princeton University and Harvard University, concurrent with developments at Bell Labs and discussions among scholars influenced by Alfred Tarski and Stephen Kleene. Early narrative threads connect to the automated reasoning communities at RAND Corporation and initiatives funded by agencies such as DARPA and NSF. Subsequent refinement and practical adoption were propelled by teams at Stanford University, Carnegie Mellon University, and industrial groups at IBM Research and Microsoft Research, leading into the era marked by conferences like CADE and SAT Competition.

Algorithm description

The method operates on formulas in conjunctive normal form and proceeds by selecting a propositional variable, assigning a truth value, applying simplification rules, and recursively searching. Core operational steps reflect techniques studied by researchers at Princeton University, Harvard University, MIT, Bell Labs, and IBM Research. Unit propagation reduces clauses when a single literal remains, a process reminiscent of work by scholars at Carnegie Mellon University and Stanford University, while pure literal elimination offers further pruning as seen in studies at University of California, Berkeley and Microsoft Research. The search employs chronological backtracking until either a complete satisfying assignment is found or contradiction is proven, echoing algorithmic paradigms explored at AT&T and Bell Labs.

Correctness and completeness

Correctness and completeness proofs derive from classical results in first-order and propositional logic developed by Alonzo Church, Kurt Gödel, Emil Post, and Alan Turing. Soundness follows because every deduction step preserves satisfiability, a property analyzed in formal methods work at Princeton University and Harvard University. Completeness is guaranteed by exhaustive search across assignments, a principle used in mechanized reasoning systems at Carnegie Mellon University and Stanford University. Countermodel construction and refutation certificates produced by the procedure played roles in verification projects at IBM and Microsoft.

Complexity and performance

The worst-case time complexity is exponential in the number of propositional variables, a limitation central to complexity theory discussions involving Stephen Cook, Richard Karp, Leonid Levin, and institutions like Princeton University and University of California, Berkeley. Practical performance depends on heuristics for variable selection and propagation; influential heuristic research arose from groups at Stanford University, Carnegie Mellon University, ETH Zurich, and industrial laboratories such as IBM Research and Microsoft Research. Modern solver architectures inspired by the original procedure integrate conflict-driven clause learning developed in research communities at MIT and EPFL, and benchmark evaluations appear in venues like SAT Competition and CADE.

Variants and extensions

Extensions include Boolean constraint propagation, non-chronological backtracking, conflict-driven clause learning, and hybrid techniques combining local search and systematic search. These advancements were pioneered by researchers affiliated with University of California, Berkeley, Cornell University, ETH Zurich, EPFL, Microsoft Research, and IBM Research. The development of clause learning and restarts links to workshops and conferences such as SAT Competition and IJCAI, and influenced tools developed at Google and Facebook for hardware verification and model checking.

Applications and implementations

Implementations of the procedure and its descendants are embedded in software used for formal verification, hardware synthesis, and combinatorial optimization, produced by organizations including IBM, Microsoft, Google, and academic groups at Stanford University, Carnegie Mellon University, and ETH Zurich. The algorithm underpins SAT-based model checkers used in projects at NASA and DARPA and appears in synthesis flows at Intel and AMD. Open-source and commercial SAT solvers drawing on these ideas are distributed by communities around SAT Competition and integrated into toolchains for teams at University of California, Berkeley and Cornell University.

Category:Algorithms