|
|
Modelling and Solving Problems Using SAT Techniques
Balyo, Tomáš ; Barták, Roman (advisor) ; Železný, Filip (referee) ; Biere, Armin (referee)
Solving planning problems via translation to satisfiability (SAT) is one of the most successful approaches to automated planning. In this thesis we describe several ways of encoding a planning problem represented in the SAS+ formalism into SAT. We review and adapt existing encoding schemes as well as introduce new original encodings. We compare the encodings by calculating upper bounds on the size of the formulas they produce as well as by running extensive experiments on benchmark problems from the 2011 International Planning Competition (IPC). In the experimental section we also compare our encodings with the state-of-the-art encodings of the planner Madagascar. The experiments show, that our techniques can outperform these state-of-the-art encodings. In the presented thesis we also deal with a special case of post-planning optimization -- elimination of redundant actions. The elimination of all redundant actions is NP-complete. We review the existing polynomial heuristic approaches and propose our own heuristic approach which can eliminate a higher number and more costly redundant actions than the existing techniques. We also propose a SAT encoding for the problem of plan redundancy which together with MaxSAT solvers allows us to solve the problem of action elimination optimally. Experiments done with...
|
|
|
Resolution-based methods for linear temporal reasoning
Suda, Martin ; Barták, Roman (advisor) ; Hoffmann, Jörg (referee) ; Biere, Armin (referee)
The aim of this thesis is to explore the potential of resolution-based methods for linear temporal reasoning. On the abstract level, this means to develop new algorithms for automated reasoning about properties of systems which evolve in time. More concretely, we will: 1) show how to adapt the superposition framework to proving theorems in propositional Linear Temporal Logic (LTL), 2) use a connection between superposition and the CDCL calculus of modern SAT solvers to come up with an efficient LTL prover, 3) specialize the previous to reachability properties and discover a close connection to Property Directed Reachability (PDR), an algorithm recently developed for model checking of hardware circuits, 4) further improve PDR by providing a new technique for enhancing clause propagation phase of the algorithm, and 5) adapt PDR to automated planning by replacing the SAT solver inside with a planning-specific procedure. We implemented the proposed ideas and provide experimental results which demonstrate their practical potential on representative benchmark sets. Our system LS4 is shown to be the strongest LTL prover currently publicly available. The mentioned enhancement of PDR substantially improves the performance of our implementation of the algorithm for hardware model checking in the multi-property...
|
|
|
Modelling and Solving Problems Using SAT Techniques
Balyo, Tomáš ; Barták, Roman (advisor) ; Železný, Filip (referee) ; Biere, Armin (referee)
Solving planning problems via translation to satisfiability (SAT) is one of the most successful approaches to automated planning. In this thesis we describe several ways of encoding a planning problem represented in the SAS+ formalism into SAT. We review and adapt existing encoding schemes as well as introduce new original encodings. We compare the encodings by calculating upper bounds on the size of the formulas they produce as well as by running extensive experiments on benchmark problems from the 2011 International Planning Competition (IPC). In the experimental section we also compare our encodings with the state-of-the-art encodings of the planner Madagascar. The experiments show, that our techniques can outperform these state-of-the-art encodings. In the presented thesis we also deal with a special case of post-planning optimization -- elimination of redundant actions. The elimination of all redundant actions is NP-complete. We review the existing polynomial heuristic approaches and propose our own heuristic approach which can eliminate a higher number and more costly redundant actions than the existing techniques. We also propose a SAT encoding for the problem of plan redundancy which together with MaxSAT solvers allows us to solve the problem of action elimination optimally. Experiments done with...
|
guest ::
