MIT16_410F10_lec11b

MIT16_410F10_lec11b - Encoding Planning Problems as...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: 7/2/11 Encoding Planning Problems as Proposi6onal Logic Sa6sfiability Sertac Karaman 16.410‐13 October 18th, 2010 Assignment •  Remember: •  Problem Set #6 Propositional Logic, due next Wednesday, October 27th. •  16:413 Project Part 1: Sat-based Activity Planner, due Wednesday, November 3rd. •  Reading –  Today: [AIMA] Chapter 10, re-read sections on SatPlan. –  Monday: Johan de Kleer and Brian C. Williams, "Diagnosing Multiple Faults," Artificial Intelligence, 32:100-117, 1987. 10/25/10 copyright Brian Williams, 2000-10 2 1 7/2/11 Planning problem •  Recall the planning problem: –  Objects •  robot1, robot2, load1, load2, room1 –  Predicates describing proper6es of objects •  (IN ?robot ?room), (HAS ?robot ?load) –  Ac6ons as means to change these proper6es •  Navigate (?robot, ?room_from, ?room_to) –  Ini6al condi6on –  Goal statement OX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXB BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX OXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX B BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX BOXBOX Image by MIT OpenCourseWare. Proposi6onal logic SAT problem •  Recall the SAT problem: –  Given a set of clauses, find an assignment to all proposi6ons to sa6sfy all the clauses. p1 ∨ ¬ p2 ∨ p 3 ¬ p1 ∨ ¬ p2 ∨ p 4 ¬ p3 ∨ p4 ∨ p 5 •  SAT solvers are very powerful. •  Can process problems with tens of thousands of variables 2 7/2/11 Encoding planning as SAT •  Idea: –  Define proposi6ons for predicates and decisions –  Encode problem descrip6on in proposi6onal logic initial state ∧ all possible action descriptions ∧ goal Encoding planning as SAT •  Ini;al condi;on •  Encode the truth of predicates: 0 0 (IN robot1 bedroom) ∧ (IN robot2 kitchen) •  Remember to include those that are false: 0 0 ¬(IN robot1 kitchen) ∧ ¬(IN robot2 bedroom) 3 7/2/11 Encoding planning as SAT •  Ac;ons •  StraighOorward approach: •  One proposi6on for each ac6on: 0 Navigate(robot1 bedroom kitchen) •  True if robot navigates from bedroom to kitchen at 6me 0 1 (IN robot1 kitchen) ⇔ 0 0 0 ((IN robot1 kitchen) ∧ ¬(Navigate(robot1, kitchen, bedroom) ∧ (IN robot1 kitchen) )) 0 0 ∨(Navigate(robot1, bedroom, kitchen) ∧ (IN robot1 bedroom) )) Robot was in the kitchen at 6me 0 and did not leave the kitchen at 6me 0. Robot was in the bedroom at 6me 0 and leR the bedroom to go to kitchen at 6me 0. Encoding planning as SAT •  Ac;ons •  What may go wrong? 0 Navigate(robot1, kitchen, bedroom) –  However, robot1 is not in the kitchen at 6me 0 ! •  PrecondiAon axioms: 0 0 Navigate(robot1, kitchen, bedroom) ⇒ (IN robot1 kitchen) 4 7/2/11 Encoding planning as SAT •  Ac;ons •  What else may go wrong? 0 Navigate(robot1, kitchen, bedroom) 0 Navigate(robot1, bedroom, livingroom) •  Ensure that one ac6on can be taken at a 6me: 0 0 ¬(Navigate(robot1, kitchen, bedroom) ∧ Navigate(robot1, bedroom, livingroom) ) Encoding planning as SAT •  Outline of the algorithm: –  Check sa6sfiability for increasing number of steps i = 1 If sa6sfiable for i steps then construct the solu6on Else i = i + 1 5 MIT OpenCourseWare http://ocw.mit.edu 16.410 / 16.413 Principles of Autonomy and Decision Making Fall 2010 For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms. ...
View Full Document

  • Fall '10
  • Prof.BrianWilliams
  • MIT OpenCourseWare, Massachusetts Institute of Technology, AirTrain Newark, OpenCourseWare, Brian C. Williams, BOXBOX

{[ snackBarMessage ]}

Ask a homework question - tutors are online