Presentation is loading. Please wait.

Presentation is loading. Please wait.

Planning based on Model Checking Dept. of Information Systems and Applied CS Bamberg University Seminar Paper Svetlana Balinova.

Similar presentations


Presentation on theme: "Planning based on Model Checking Dept. of Information Systems and Applied CS Bamberg University Seminar Paper Svetlana Balinova."— Presentation transcript:

1 Planning based on Model Checking Dept. of Information Systems and Applied CS Bamberg University Seminar Paper Svetlana Balinova

2 1. Introduction 2. Explicit Model Checking 3. Temporal Logic 4. Symbolic Model Checking 5. Binary Decision Diagrams 6. Planning for Reachability Goals Outline

3 „ Model Checking is an automatic technique for verifying correctness properties of safety-critical reactive systems “ 1. Introduction Classical planningPlanning under uncertainty DeterminismNondeterminsm Full observabilityPartial observability Reachability goalsExtended Goals 2 kinds of correctness properties: Safety Liveness

4 State space: a system, implemented as Kripke Structure Verification: build a computation tree for all possible paths within a System 2. Explicit Model Checking

5 green signal Warning system for planes signal, evasion recommendation

6 Properties to be checked must be formalized in a temporal logic. A temporal logic provides operators which represent time dependences. Operators: F (in the future), X (next time), G (globaly) Path quantifiers: A (always), E (exist) CTL (Computation Tree Logic) – branching time. LTL (Linear-Time Temporal Logic) – linear time. Validity of LTL and CTL formulas: Model Checking Algorithms. 3. Temporal Logic

7 Examples: φφ φ LTL formula: F φ CTL formula: AX φ ……………………. ……..

8 Faces the state explosion problem of explicit state Model Checking. Exploring sets of states, rather than single states. In order to represent a model checking problem symbolically, we need to represent symbolically: 4. Symbolic Model Checking the sets of states of a Kripke Structure, its transition relation, and the model checking algorithms.

9 Symbolic Representation of Sets of States: A vector x of Boolean variables where each variable corresponds to a an atomic propostion in P. x = { green, signal, evasiion recommendation} A state s is represented with a formula ξ(s) on the propositions: ξ(s o ) = green, ¬signal, ¬evasion recommendation ξ(s 1 ) = ¬green, signal, ¬evasion recommendation ξ(s 2 ) = ¬green, signal, evasion recommendation 4. Symbolic Model Checking A set of states QS represented symbolically as: 

10 Symbolic Representation of Transition Relations A vector of state variables x = and a further vector of next state variables x‘ = x‘ = { green’, signal’, evasion recommendation’} A transition ξ( ) encoded as: ξ( ) = ξ( ), ξ‘( ) ξ( ) = (green, ¬signal, ¬evasion recommendation), (¬green‘, signal’, ¬evasion recommendation’) Transition relation R represented symbolically as: ξ(R) = V ξ(r) ξ(R) = V ξ(r) 4. Symbolic Model Checking r є R

11 Symbolic Representation of Model Checking Algorithms Replace each function call with the symbolic counterpart. Cast the operations on sets into the corresponding operations on propositional formulas. 4. Symbolic Model Checking

12 An efficient approach for manipulation of Boolean formualas. A binary decision diagram represents a Boolean function as rooted, directed acyclic graph. Each nonterminal vertex v is labeled by a variable var(v) and has ars directed toward two children: lo(v) and hi(v). In the first case the variable is assigned 0(- - -) and in the second 1( ). Each terminal vertex is labeled 0 or 1 A Boolean function may be represented by a truth table, binary decision tree etc. A tree is said to be ordered if the variables always occur in the same order along any path from root to leaf. 5.Binary Decision Diagrams

13 Truth Table and Decision Tree Representations of a Boolean Function. A dashed (solid) tree branch denotes the case where the decision variable is 0 (1). Example:

14 Reduction of decision tree to OBDD 1. Remove Duplicate Terminals: Eliminate all but one terminal vertex with a given label and redirect all arcs into the eliminated vertices to the remaining one. 2. Remove Duplicate Nonterminals: If nonterminal vertices u and v have var(u)=var(v), lo(u)=lo(v), and hi(u)=hi(v), then eliminate one of the two vertices and redirect all incoming arcs to the other vertex. 3. Remove Redundant Tests: If nonterminal vertex v has lo(v)=hi(v), then eliminate v and redirect all incoming arcs to lo(v). 5.Binary Decision Diagrams

15 Applying the three reduction rules to the tree of the last example yields the canonical representation of the function as an OBDD.

16 Rechability goals: Goals are sets of states, i.e., the objective is to build a plan that leads to one of the goal states. A planning domain is a nondeterministic state-transition system Σ= (S, A,  ), where: S is a finite set of states A is a finite set of actions  : S  A  2 s is the state-transition function. 6. Planning for Reachability Goals

17 Example: nondeteministic state-transition system for a simplified DWR (dock-worker-robots) domain

18 6. Planning for Reachability Goals A plan is a policy, i.e. a function that maps states into actions. A policy π for a planning domain Σ= (S, A,  ) is a set of pairs (s, a) such that (s, a) є A(s). π1 = {(s1, move(r1,l1,l2)), (s2, move(r1,l2,l3)), (s3, move(r1,l3,l4))} π2 = {(s1, move(r1,l1,l2)), (s2, move(r1,l2,l3)), (s3, move(r1,l3,l4)), (s5, move(r1,l3,l4))} π3 = {(s1, move(r1,l1,l4))} Policies for the domain in the previous examle:

19 We represent the execution of a policy in a planning domain with an execution structure, i.e., a directed graph in which the nodes are all of the states of the domain that can be reached by executing actions in the policy, and the arcs represent possible state transitions caused by actions in the policy. 6. Planning for Reachability Goals π1π1 π2π2 π3π3

20 A planning problem is a triple (Σ, S o, S g ), where Σ= (S, A,  ) is a planning domain, S o  S is a set of initial states, and S g  S is a set of goal states. Types of solutions for a planning problem: weak solutions strong solutions strong cyclic solutions Planning algorithms – designed to work on sets of states, thus taking advantage of the BDD-based symbolic Model Checking. 6. Planning for Reachability Goals

21 Thank you for your attention!


Download ppt "Planning based on Model Checking Dept. of Information Systems and Applied CS Bamberg University Seminar Paper Svetlana Balinova."

Similar presentations


Ads by Google