1 / 13

Project Topics: MCI 2007.1

Project Topics: MCI 2007.1. Jacques Robin. Topics Supervised by Prof. Jacques. Developing an ontology and component framework of search algorithms Top-level ontology classes derived from Russell & Norvig and Dechter Leaves of the ontology: to work with a Constraint Handling Rules (CHR)

paxton
Download Presentation

Project Topics: MCI 2007.1

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Project Topics: MCI 2007.1 Jacques Robin

  2. Topics Supervised by Prof. Jacques • Developing an ontology and component framework of search algorithms • Top-level ontology classes derived from Russell & Norvig and Dechter • Leaves of the ontology: • to work with a Constraint Handling Rules (CHR) • to include variations of conflict-directed backjumping for complete global search • to include variations of min-conflict for incomplete, local, scalable search • Using UML knowledge representation and transformation rules from UML Components to Java OSGi Components • Incrementally developing multi-agent simulation • Starting from simplest simulation of penalty shot • Using UML as knowledge layer representation language • Using Java and CHR as implementation layer representation language

  3. Topics Supervised by Prof. Fred • ??

  4. <<interface>> StateSpaceSearch +gSearch(StateSpaceSearchPb):SearchSolution <<component>> PruningHeuristic +prune(Node):Node[*] <<component>> BtStrategy +bt(Node):Node <<interface>> Cost2GoalHeuristic +estimCost2Goal(Node):Real <<interface>> PruningHeuristic +prune(Node):Node[*] <<interface>> BtStrategy +bt(Node):Node <<component>> Cost2GoalHeuristic +estimCost2Goal(Node):Real <<component>> StateSpaceSearch +gSearch(StateSpaceSearchPb):SearchSolution <<interface>> ExpandStrategy +choose(Fringe):Node <<component>> ExpandStrategy +choose(Fringe):Node StateSpaceSearchPb +fullStateFornulation; Boolean +suc(State,AgentAction):State Path +/cost:Real Framework Top-Level 1..* AgentAction +name:String +cost:Real <<uses>> <<uses>> <<uses>> <<uses>> State +full:Boolean +goal:Boolean +initial:Boolean 2..* parent models child * Node +/expanded: Boolean +/root: Boolean +/visited: Integer Fringe * NodeSolution {ordered} * SearchSolution PathSolution

  5. Prof. Jacques’ Search Topic VariableChoice Heuristic SearchProblem SearchAlgo PartialStateFormulation SearchProblem FullStateFormulationSearchProblem GlobalSearchAlgo LocalSearchAlgo CSPSearchProblem CSPSearchAlgo FDCSP SearchProblem FDCSP SearchAlgo ValueChoice Heuristic PartialStateFormulation FDCSPSearchProblem FullStateFormulation FDCSPSearchProblem GlobalFDCSP SearchAlgo LocalFDCSP SearchAlgo Backtracking Heuristic CDBJ Min-Confllict

  6. Tasks • Model UML2/OCL2 hierarchy of abstract and concrete specializations of StateSpaceSearchProblem (2 students) • including concrete classes and instances of: • 8 queens, CSP backjumping slides map coloring • Model UML2/OCL2 hierarchy of abstract and concrete specializations of CoastToGoalHeuristic (except for CSP problem, fully problem dependent) (same 2 students than 1) • Model UML2/OCL2 hierarchy of abstract and concrete specializations of ExpandStrategy and PruningHeuristic (Fúlvio) • Model UML2/OCL2 hierarchy of abstract and concrete specializations of BtStrategy (Zé Carlos, Renan) • Model UML2/OCL2 hierarchy of abstract and concrete specializations of StateSpaceSearch as assembly of abstract and concrete specializations ExpandStrategy, BtStrategy, PruningHeuristic, CoastToGoalHeuristic (Carlos, Alexandre) • Model UML2/OCL2 or other technology search visualization GUI(Joabe, speak to Luiz Lacerda, luiz.francisco.lacerda@gmail.com, about his UML2 Profile for GUI Modeling)

  7. Tasks • Implementation OSGi Java, tests JUnit hierarchy of abstract and concrete specializations of StateSpaceSearchProblem (2 students) • including concrete classes and instances of: • 8 queens, CSP backjumping slides map coloring • Implementation OSGi Java, tests JUnit hierarchy of abstract and concrete specializations of CoastToGoalHeuristic (except for CSP problem, fully problem dependent) (same 2 students than 1) • Implementation OSGi Java, tests JUnit hierarchy of abstract and concrete specializations of ExpandStrategy and PruningHeuristic (Fúlvio) • Implementation OSGi Java, tests JUnit hierarchy of abstract and concrete specializations of BtStrategy (Zé Carlos, Renan) • Implementation OSGi Java, tests JUnit hierarchy of abstract and concrete specializations of StateSpaceSearch as assembly of abstract and concrete specializations ExpandStrategy, BtStrategy, PruningHeuristic, CoastToGoalHeuristic (Carlos, Alexandre) • Implementation OSGi Java, tests JUnit or other technology search visualization GUI(Joabe, speak to Luiz Lacerda, luiz.francisco.lacerda@gmail.com, about his UML2 Profile for GUI Modeling)

  8. Scope Search Problems • Priority1: • FullStateStateFormulation, PartialStateFormulation • CSPFullStateStateFormulation, CSPPartialStateFormulation • N-queens as FullStateStateFormulation • N-queens as PartialStateFormulation • 8-queens as FullStateStateFormulation • 8-queens as PartialStateFormulation • MapColoring as FullStateStateFormulation • R1-R7 MapColoring as FullStateStateFormulation • Priority 2: • PathSolutionProblem • ShortestPathBetween2Cities • Romenia

  9. Scope Expand and Pruning Strategies • General StateSpaceSearch • Priority 1: • For FullStateFormulationProblems: Depth-first, Backtracking • For PartialStateFormulationProblems: min-conflict • Priority 2: Uniform cost search, A* • Priority 3: Breadth-first, iterative deepening, RBFS • CSPSearch: • Priority 1: • Variable ordering: Degree Heuristic • Value ordering: Least Constraining Value • Pruning: forward checking • Priority 2: • Variable ordering: Minimum Remaining Value • Arc consistency

  10. Scope Backtrack Strategies • Chronological backtracking • Conflict-directed backjumping

  11. Time Table • 09-13/07: Version 1.0 of first half of model • 16-20/07: Version 2.0 of first half of model • 30/07-03/08: Version 1.0 of first half implementation and 1.0 of second half of model • 13/08-17/08: Version 1.1 of first half implementation and 1.0 of second half implementation and integration tests • 22/08: Final report

  12. Topic 2 Starting Point: Simplest Possible Multi-Agent Simulation percept Simulation Agent Shooter Agent Keeper Agent action action percept gameOver, goal action(s,legs,shoot(2)) action(k,legs,move(right)) actions(k,legs,move(right)) action(k,hands,grab(yes)) action(s,legs,shoot(3)) action(k,legs,move(right)) gameOver, nogoal

  13. Topic 2 Possible Task Division • Simulation Agent: • Reasoning • Simulation Visualization: • Agent Reasoning Explanation Visualization • Shooter Agent: • Reasoning • Simulation Visualization: • Agent Reasoning Explanation Visualization • Keeper Agent • Reasoning • Simulation Visualization: • Agent Reasoning Explanation Visualization

More Related