CS 151: Intelligent Agents, Problem Formulation and Search

Transcription

1

2 CS 151: Intelligent Agents, Problem Formulation and Search

3 How do we make a computer "smart?" Computer, clean the house! Um OK?? This one's got no chance

4 How do we represent this problem? Hmmm where to begin? We want the vacuum to "clean the house." What does this mean? What's involved for the vacuum cleaner? How can we formalize this problem a bit?

5 Vacuum Cleaner World Percepts: location and contents, e.g., [A,Dirty] Actions: Left, Right, Suck, NoOp

6 Rational Agents Rational Agent = An agent (program) that does the "right" thing, given its goals, its abilities, what it perceives of its environment and its prior knowledge What do we need to know in order to decide if the vacuum cleaner is rational?

7 What does the agent DO? Our goal as AI programmers is develop agents that behave rationally. This means we must specify what the agent does given: Its goals Its precepts (what is perceives) Its possible actions Its prior knowledge This plan of action is the agent's policy AI Programming = Policy Design and Implementation

8 Defining a Policy How might we define a policy for the vacuum agent above if the goal is to clean both rooms? (Hint: there's more than one right answer) What do we need to know? What do we need to watch out for?

9 Techniques for Implementing Policies Search Reasoning with knowledge and uncertainty Reasoning with Utility At the core of this class will be several techniques for Policy design and implementation Learning

10 Vacuum world Search-based agent 1. Formulate problem and goal 2. Search for a sequence of actions that will lead to the goal (the policy) 3. Execute the actions one at a time

11 Vacuum world Search-based agent 1. Formulate problem and goal 2. Search for a sequence of actions that will lead to the goal (the policy) 3. Execute the actions one at a time Well-defined problem: (State space) Initial state Goal test Actions/Successor function Path cost

12 Vacuum world States: Shown above Actions: Move (R or L), Suck, (NoOp) Successor function given by state graph (next slide) Goal test: In state 7 or 8? Path cost: +1 for each move and each suck

13 Vacuum world state space graph

14 Search problem types Deterministic, fully observable single-state problem Agent knows exactly which state it will be in; solution is a sequence of actions Non-observable sensorless (conformant) problem Agent may have no idea where it is; solution is a sequence Nondeterministic and/or partially observable contingency problem percepts provide new information about current state often interleave search, execution Unknown state space exploration problem

15 Example: vacuum world Single-state, start in #5. Solution?

16 Example: vacuum world Sensorless, start in {1,2,3,4,5,6,7,8} e.g., Right goes to {2,4,6,8} Solution?

17 Example: Vacuum world Contingency Nondeterministic: Suck may dirty a clean carpet Partially observable: location, dirt at current location. Percept: [L, Clean], i.e., start in #5 or #7 Solution?

18 Example #2: Navigation I'm at school and I want to go to Madison Seattle 2403 Madison Chicago 830 New York Ontario Dallas How can you formulate this problem?

19 Example from book Romania

20 Example #3: The 8-puzzle states? actions? goal test? path cost?

21 Vacuum world Search-based agent 1. Formulate problem and goal 2. Search for a sequence of actions that will lead to the goal (the policy) 3. Execute the actions one at a time

22 Finding the path: Tree search algorithms Basic idea: offline, simulated exploration of state space by generating successors of already-explored states (a.k.a. expanding states) def TreeSearch(problem, strategy): initialize search tree using information in the problem while true: if there are no candidates for expansion, return failure choose a leaf node for expansion according to strategy if node contains goal state, return solution else expand node and add resulting nodes to search tree

23 Careful! states vs. nodes A state is a (representation of) a physical configuration A node is a data structure constituting part of a search tree includes state, parent node, action, path cost g(x), depth The Expand function creates new nodes, filling in the various fields and using the SuccessorFn of the problem to create the corresponding states.

24 Tree search example I'm at school and I want to go to Madison Seattle 2403 Madison Chicago 830 New York Ontario Dallas

25 Implementation: general tree search

26 Search strategies A search strategy is defined by picking the order of node expansion How to evaluate a strategy?

27 Uninformed search strategies Uninformed search strategies use only the information available in the problem definition Depth First Search Breadth First Search

28 Breadth-first search Expand shallowest unexpanded node Implementation: fringe is a FIFO queue, i.e., new successors go at end

29 Breadth-first search Expand shallowest unexpanded node Implementation: fringe is a FIFO queue, i.e., new successors go at end

30 Breadth-first search Expand shallowest unexpanded node Implementation: fringe is a FIFO queue, i.e., new successors go at end

31 Breadth-first search Expand shallowest unexpanded node Implementation: fringe is a FIFO queue, i.e., new successors go at end

32 Depth-first search Expand deepest unexpanded node Implementation: fringe = LIFO queue, i.e., put successors at front

33 Depth-first search Expand deepest unexpanded node Implementation: fringe = LIFO queue, i.e., put successors at front

34 Depth-first search Expand deepest unexpanded node Implementation: fringe = LIFO queue, i.e., put successors at front

35 Depth-first search Expand deepest unexpanded node Implementation: fringe = LIFO queue, i.e., put successors at front

36 Depth-first search Expand deepest unexpanded node Implementation: fringe = LIFO queue, i.e., put successors at front

37 Depth-first search Expand deepest unexpanded node Implementation: fringe = LIFO queue, i.e., put successors at front

38 Depth-first search Expand deepest unexpanded node Implementation: fringe = LIFO queue, i.e., put successors at front

39 Depth-first search Expand deepest unexpanded node Implementation: fringe = LIFO queue, i.e., put successors at front

40 Depth-first search Expand deepest unexpanded node Implementation: fringe = LIFO queue, i.e., put successors at front

41 Depth-first search Expand deepest unexpanded node Implementation: fringe = LIFO queue, i.e., put successors at front

42 Depth-first search Expand deepest unexpanded node Implementation: fringe = LIFO queue, i.e., put successors at front

43 Depth-first search Expand deepest unexpanded node Implementation: fringe = LIFO queue, i.e., put successors at front

44 Activity Analyze DFS and BFS according to the criteria time, space, optimal, complete (for time and space, analyze in terms of b, d, and m; for complete and optimal - simply YES or NO) Which strategy would you use and why? Brainstorm improvements to DFS and BFS

45 Properties of breadth-first search Complete? Time? Space? Optimal?

46 Time and Memory requirements for BFS Depth Nodes Time Memory sec 1 MB 4 111, sec 106 MB min 10 GB hours 1 terabyte days 101 terabytes years 10 petabytes ,523 years 1 exabyte BFS with b=10, 10,000 nodes/sec; 10 bytes/node

47 Uniform-cost search Expand least-cost unexpanded node Implementation: fringe = queue ordered by path cost Equivalent to breadth-first if step costs all equal Complete? Yes, if step cost ε Time? # of nodes with g cost of optimal solution, O(b ceiling(c*/ ε) ) where C * is the cost of the optimal solution Space? # of nodes with g cost of optimal solution, O(b ceiling(c*/ ε) ) Optimal? Yes nodes expanded in increasing order of g(n)

48 Properties of depth-first search Complete? Time? Space? Optimal?

49 Next Time: More Search Depth-limited search Iterative deepening search Smarter search