using cp when you don t know cp
play

Using CP When You Don't Know CP Christian Bessiere LIRMM (CNRS/U. - PowerPoint PPT Presentation

Oct 02, 2023 •244 likes •693 views

Using CP When You Don't Know CP Christian Bessiere LIRMM (CNRS/U. Montpellier) An illustrative example 5-rooms flat (bedroom, bath, kitchen, sitting, dining) on a 6-room pattern The pattern: Constraints of the builder: north north- north

  1. Using CP When You Don't Know CP Christian Bessiere LIRMM (CNRS/U. Montpellier)

  2. An illustrative example 5-rooms flat (bedroom, bath, kitchen, sitting, dining) on a 6-room pattern The pattern: Constraints of the builder: north north- north – Kitchen and dining must be -west east linked – Bath and kitchen must have a south- south- common wall south west east – Bath must be far from sitting – Sitting and dining form a single room

  3. Problem • How to propose all possible plans?  a constraint network that encodes the constraints of the builder

  4. Library of constraints • Constraints : nw n ne – X ≠ Y, X = Y se sw s – Next (X,Y) = { (nw,n),(nw,sw),(n,nw),(n,ne),(n,s), (ne,n),(ne,se),(sw,nw),(sw,s),(s,sw), (s,n),(s,se),(se,s),(se,ne) } – Far (X,Y) = { (nw,ne),(nw,s),(nw,se),(n,sw),(n,se), (ne,nw),(ne,sw),(ne,s),(sw,n),(sw,ne), (so,se),(s,nw),(s,ne),(se,nw),(se,n),(se,sw) }

  5. A possible viewpoint (variables, domains) • Variables : nw n ne – B (bedroom), se sw s – W (washroom), – K (kitchen), – S (sitting), – D (dining) • Domains : {nw,n,ne,sw,s,se}

  6. A constraint network K B D next nw,n,ne,sw,s,se nw,n,ne,sw,s,se nw,n,ne,sw,s,se next next W nw,n,ne,sw,s,se S nw,n,ne,sw,s,se far Alldiff (K,W,D,S,B) K W Bedroom: east a solution Client wishes: S D B Sitting: south

  7. Constraint Programming modelling Problem Variables Domains ??? Constraints Solution solving

  8. Modelling (“ it’s an art, not a science ”) • In the 80s, it was considered as trivial – Zebra problem (Lewis Carroll) or random problems • But on “real” problems: – Which variables ? Which domains ? – Which constraints for encoding the problem? • And efficiency? – Which constraints for speeding up the solver? • Global constraints, symmetries…  All is in the expertise of the user

  9. If you’re not an expert? 1. Choice of variables/domains 2. Constraint acquisition 3. Improve a basic model

  10. Choice of variables/domains (viewpoints) • From historical data (former solutions) • Solutions described in tables (flat data) Room Position Room Position Dining nw D S Room Position Sitting nw Kitchen n W ash nw W S K D C Kitchen n Sitting K itchen ne n Bedroom ne B W K K S Bedroom B edroom sw sw SàM Wash s D ining s B D S Wash se S itting se

  11. Extract viewpoints X Wash ∈ {nw,n,ne,sw,s,se} Room, position …. Room Position X Sitting ∈ {nw,n,ne,sw,s,se} ∅ Room Position Wash nw Kitchen n Bedroom sw Dining s Sitting se

  12. Extract viewpoints Room, position • Two viewpoints: – X B ,…,X S ∈ {nw,n,ne,sw,s,se} Room Position – X nw ,…,X se ∈ {W,B,K,D,S, ∇ } ∅ • Trivial viewpoints: – X 1 ,…,X 5 ∈ {B-nw,B-n,B-sw,…, Room Position S-s,S-se} wash nw kitchen n – X B-nw ,…,X S-se ∈ {0,1} bedroom sw dining s sitting se

  13. Connect viewpoints • VP1: X B ,…,X S ∈ {nw,n,ne,sw,s,se} • VP2: X nw ,…,X se ∈ {B,W,K,D,S, ∇ } • Channelling constraints: – X B = nw ↔ X nw = B  “nw” is taken at most once in VP1  alldiff (X B ,…,X S ) is a constraint in VP1 [like in Law,Lee,Smith07]

  14. Application: sudoku L C V L1 C4 3 X LC =V L1 C5 1 L2 C1 3 X LV =C L2 C3 4 L3 C4 2 X CV =L L3 C9 8 … … … Alldiffs learned for free

  15. Connect viewpoints • We can derive more than just alldiff • Cardinality constraints can be detected • Example: a timetabling in which 3 math courses are given  one of the viewpoints will contain 3 variables representing these 3 courses  In all other viewpoints, we can put a cardinality constraint forcing value “math” to be taken 3 times

  16. If you’re not an expert? • Choice of variables/domains • Constraint Acquisition – Space of networks – Redundancy – Queries • Improve a basic model

  17. Acquire constraints • The user doesn’t know how to specify constraints • She knows how to discriminate solutions from non-solutions – Ex: valid flat vs invalid flat  Use of machine learning techniques – Interaction by examples (positive e+ or negative e- ) – Acquisition of a network describing the problem

  18. Space of possible networks ? ? X K X D Some negative accepted ? ? X B ? ? ? X W X S ? • Language : ? → { =, ≠, next , far } • Bias : Some positive rejected X S =X W ; next (X S ,X B );… …; X K ≠X D ; far (X K ,X D )

  19. Compact SAT encoding • A SAT formula K representing all Some negative accepted possible networks: – Each constraint c i → a literal b i – Models( K ) = version space – Example e- rejected by {c i , c j , c k } → a clause ( b i ∨ b j ∨ b k ) – Example e+ rejected by c i → a clauses ( ¬ b i ) • m ∈ models( K ) Some positive rejected ⇒ ϕ (m) = { c i | m ( b i )=1} accepts all positive examples and rejects all negative examples

  20. Reduce the space e + 1 B K C(X K ,X D ): ≠ , = , far , next S D W e - K W C(X D ,X S ): ≠ , = , far , next 2 S D B e + C(X K ,X S ): ≠ , = , far , next B 3 K D S W e - 4 next (X D ,X S ) ∨ far (X K ,X S ) S B K D M W

  21. Redundancy K M • Constraints are not S independent • “ next (X K ,X D ) ∧ next (X D ,X S ) ⇒ far (X K ,X S )” • See local consistencies • It’s different from attribute-value learning

  22. Redundancy • Redundancy prevents convergence  a set R of redundancy rules: K alldiff (X 1 ,…,X n ) ⇒ X i ≠X j , ∀ i,j M S next (X K ,X D ) ∧ next (X D ,X S ) ⇒ far (X K ,X S ) • In K we already have: – next (X D ,X S ) ∨ far(X K ,X S ) – next (X K ,X D ) • So, from K + R we deduce far (X K ,X S ) • Version space = Models( K + R ) – Good properties when R is complete

  23. Queries (active learning) • Examples often lead to little new information (eg, negative plan with kitchen far from dining) • The system will propose examples (queries) to speed up convergence • Example e rejected by k constraints from the space – e positive ⇒ k constraints discarded from the space – e negative ⇒ a clause of size k • Good query = example which reduces the space as much as possible whatever the answer

  24. Queries K + R b 1 ∨ … ∨ b k • Negative example e1 :  cl e1 = b 1 ∨ … ∨ b k ∈ K + R – find m ∈ models( K + R ) such that a single literal b i in cl e1 is false m contains ¬ b i – find e2 ∈ sol( ϕ (m)) : → e2 violates only constraint c i ϕ (m)  b i or ¬ b i will go in K • If sol( ϕ (m) )= ∅ : any conflict-set is a e2 ∈ sol( ϕ (m) ) new redundancy rule  quick convergence Query: “ e2” ?

  25. An example of constraint acquisition in robotics (by Mathias Paulin) • The goal is to automate the burden of implementing elementary actions of a robot • Elementary actions are usually implemented by hand by engineers (complex physic laws, kinetic momentum, derivative equations, etc.)

  26. No need for a user • Instead of interacting with a user, classification of examples will be done by a run of the robot with given values of its sensorimotor actuators • If the action has correctly performed, this is positive • With expensive humanoid robots, a simulator allows easy classification without actually running the robot

  27. Elementary actions • Each action has variables representing – the observed world before the action, – the power applied to each actuator – the world after the action • Constraint acquisition will learn a constraint network on these variables such that its solutions are valid actions

  28. Planning a task • The overall goal is to build a plan composed of elementary actions • The planning problem is solved by a CP solver • It is convenient to encode actions as sub-CSPs

  29. Tribot Mindstorms NXT • 3 motors • 4 sensors • 5 elementary actions to combine • Discretization of variables

  30. Experiment Experiment • Modelling by CONACQ • Conacq generates a CHOCO model used by CSP-Plan [Lopez2003] ⇒ Objective : catch the mug!

  31. If you’re not an expert? • Choice of variables/domains • Constraint acquisition • Improve the basic model

  32. Improve the model modelling • Basic model M1 : Problem Variables BT-search + propagation solve(M1) ≈ ∞ Domains Constraints Solution  Experts add implicit solving constraints that increase constraint propagation The globalest is the best • An implicit constraint doesn’t change the set of solutions  We will learn implicit global constraints

  33. Implicit global constraints X 1 … X n sol(M1): • Model M1: 112345 at most two 1 per solution 332223 551554 124135 • M1+{ card [#1 ≤ 2](X 1 ..X n )}: ….. same solutions as M1 • But solve(M1+ card ) is faster than solve(M1) Card[..]+card[..]+card[..] = gcc [P] gcc = propagation with a flow

  34. Learn parameters P of gcc [P](X 1 ..X n ) Very hard Very easy relax M1 M2 M1+ Sol(M1) Sol(M2) gcc [P’](X 1 ..X n ) gcc [P’](X 1 ..X n ) Easy gcc [P](X 1 ..X n )

  35. Example: Task allocation • Projects to be assigned to students while minimising disappointment • Model M1 designed by some of the students (2h of courses on CP) : • optimize (M1) > 12h

Recommend

More recommend