Planning and reasoning: projects
Students may propose their own projects, but they are subject to the
professor's approval. Projects already started without this approval may still
be rejected.
Suggested projects:
- planning a sequence of software package installations, based on their
dependencies
- syntax checker for PDDL: currently, Fast-Downward just terminates printing
a backtrace when the input is syntactically incorrect; the aim of this project
is to build a system to check the validity of a PDDL file; optionally, this
module is so that it it could be integrated into Fast-Downward; the system
could also make some preliminary simplifications such as merging actions,
removing unexecutable actions, etc.
- analysis of the selection mechanisms in portfolio-based planning systems;
this project requires studying the literature on portfolio-based planning
systems and writing an essay on the methods these systems use for selecting the
planning algorithm
- library of examples of PDDL, showing in increasing order of complexity the
features of the language; also supply a graphical representation of the domains
- deontic logic
- progress meter for DPLL
- privacy preserving planning
- every path from a point to another must cross a line (or a broken line) in
the plane; this is a sort of landmark, like in domain-independent heuristics;
investigate previous explotiations of this idea, or propose how it could be
exploited
- use of a domain-dependent heuristics in fast-downward
- unsharp turns in a given path in the plane: check literature, propose ideas
- goal recognition: tecniques, applications
- extensions of LRTA*
- guarded logic
- multi-agent path planning
- soft constraints on plan trajectories: formulation in PDDL and solving
algorithms
- path planning with dynamic obstacles
- backtracking with local search: implement backtracking using local search
to determine the next branching literal (for example, the variable that gives
the two simplified problems with as few satisfiable clauses as possible)
- compressed path databases for path planning
- μ-calculus: definition, applications and algorithms
- recent algorithms for propositional model counting
- beam search and its applications to planning
- logistics planning
- hyper-heuristics
Instructions about projects based on the available literature:
- this is not an English course; I do not care about grammar mistakes as long
as the prose is comprehenisble;
- do not copy text from web sites or articles;
- the aim of the project is that the student gets a basic understanding of
the given topic;
- the aim of the report is to show that the student has obtained that
understanding;
- the project is evaluated from the report;
- therefore, the report must a. be original (no copied text); b. show that
the student has a grasp of the general picture; c. show that the student's
understanding of the technical details is sufficient for giving a sense to the
general picture;
- as an example, a beautifully formatted report that is a patchwork of random
formulae and grammatically perfect sentences copied from articles does not pass
- as an example, a report that provides a novel way of explaining the topic,
with technical details provided exactly when needed, will get an high grade
even if poorly formatted and full of grammar mistakes