Description
Question 1: Search Algorithms [50] Little Red Riding Hood wants to go to her grandmother’s house in the forest to give her some cake. The forest is a dangerous place where the Big Bad Wolf resides. Little Red Riding Hood’s mother has instructed her not to venture into the forest and to only go along the road. The following 5×5 grid shows the area that Little Red Riding Hood needs to traverse. She starts from square A (denoted by R). Her grandmother’s house is located at square N (denoted by G). The forests are denoted by F (on squares F, I, M, R, U and Y). Your task is to help Little Red Riding Hood get to her grandmother’s house. Assume that she can only move in four directions namely left, down, right and up. She cannot travel diagonally and will not go into the forests. Also assume that the successor function will cause legal moves to be examined in an counter-clockwise order: left; down; right; up. Note that not all of these moves may be possible for a given square (squares with forest on them will never be examined). A B C D E R F G H I J F F K L M N O F G P Q R S T F U V W X Y F F a) [10] Using Depth-First Search, list the squares in the order they are expanded (including the goal node if it is found). Square A is expanded first (hint: State B will be examined next). Assume cycle checking is done so that a node is not generated in the search tree if the grid position already occurs on the path from this node back to the root node (i.e., Path Checking DFS). Write down the list of states you expanded in the order they are expanded. Write down the solution path found (if any), or explain why no solution is found. b) [10] Using Breadth First Search write down the list of states you expanded in the order they are expanded (until the goal node is reached). Use the same cycle checking as in the previous question. c) [10] Using Iterative Deepening Search, draw the trees built at each depth until a solution is reached. Use the same cycle checking as in the previous questions. d) [10] Consider the heuristic function h(n) which is the Manhattan distance between a given square and the goal square. Is h(n) an admissible heuristic? e) [10] Perform A* search using the heuristic function h(n). In the case of ties, expand states in alphabetical order. List each square in the order they are added to the OPEN list, and mark it with f(n) = g(n) + h(n) (show f, g and h separately). When expanded (including square N), label a state with a number indicating when it was expanded (square A should be marked 1). Highlight the solution path found (if any), or explain why no solution is found. CS 540-1 Fall 2014 Question 2: Programming Part: A* Search[50] In this question, your task is to implement A* search algorithm to solve a game problem, the rotation game. Moreover, you will be asked to try different heuristic functions in you A* algorithm to understand the performance difference between them. Game description The rotation game uses a # shaped board, which can hold 24 pieces of square blocks (see Fig.1). The blocks are marked with symbols 1, 2 and 3, with exactly 8 pieces of each kind. Initially, the blocks are placed on the board randomly. Your task is to move the blocks so that the eight blocks placed in the center square have the same marking symbol. There is only one type of valid move, which is to rotate one of the four lines, each consisting of seven blocks. That is, six blocks in the line are moved towards the head by one block and the head block is moved to the end of the line. The eight possible moves are marked with capital letters A to H. Figure 1 illustrates two consecutive moves, move A and move C from some initial state. Figure 1: The Rotation Game. The input file will comprise of only one line, which contains 24 digits corresponding to the symbols of the blocks in the initial state. The rows of blocks are listed from top to bottom. For each row the blocks are listed from left to right. For each input file, your program should apply a specified heuristic function, and output three corresponding results for the goal state with the minimum steps of moves. The first result is the state for this goal. The second result is the operation sequence with letters ranging from ’A’ to ’H’, which represents the moves needed to reach this goal. If no moves are needed, you should output ”No moves needed” instead. The third result is the number of popped out states from the queue (including the goal itself). Notice that there could be multiple goal states with the minimum steps of moves, which leads to uncertainty in the grading process. In light of it, we will introduce the way to resolve this uncertainty in the next subsection. The following shows a sample input and output for the heuristic function given by Eq.1, and it corresponds to the board game shown in Fig. 1. CS 540-1 Fall 2014 Sample Input 111132323132231222312133 Sample Output 1 1 3 1 3322231 2 2 3122231 3 1 1 3 AC 3 Methods to implement In this programming question, you are only required to implement four methods of the class AStarSearchImpl, which are showed as follows: public class AStarSearchImpl implements AStarSearch { public SearchResult search(String initConfig, int modeFlag); public boolean checkGoal(String config); public String move(String config, char op); public int getHeuristicCost(String config, int modeFlag); } In detail, you should implement the A* algorithm strictly according to the pseudo code shown in Alg.1, which is primarily based on the codes in Prof. Zhu’s slides but with more details. As mentioned before, there might be several goal states with the minimum operation length, so there are some points that must be mentioned in order to ensure the final answer to be unique: 1. In line 3, it’s possible that there might exist multiple nodes with the same least f value. In this situation, you should always pop the node with the least f value and smaller dictionary order with respect to node’s operation sequence. In fact, we have already written a comparator function for this comparison criteria, as seen in State.java, so that you can use it to build the priority queue directly. 2. In line 8, you should generate the successor of n from operation A to operation H. 3. In line 16, you shouldn’t add an extra copy of n 0 into OPEN. Combined with line 18, it implies that at any moment, there will only at most one copy for any possible node. To implement the priority queue OPEN, we recommend (but not require) you to use the PriorityQueue in java with the comparator defined in State.java. Unfortunately, PriorityQueue in java does not support modification of its elements directly, so in order to implement line 16, you have to first use PriorityQueue.remove to remove the element in the queue with n 0 , then add a new element with n 0 and new values into the queue. To use the function remove of the PriorityQueue, you need to override the equals function for the class of elements contained in the queue. Like the comparator function, we have also finished this part for you in State.java. However, be careful that this step does not increase the third result by one, namely the number of states popped out from the queue. Finally, method search should return the result of class SearchResult, which comprises the final state, operation sequence and number of states popped out from the queue as mentioned. For other parts of programming, you also have to implement the methods checkGoal, move and getHeuristicCost, whose prototypes are well commented in AStarSearch.java. Also to reduce your CS 540-1 Fall 2014 programming burden, we have written I/O parts for you, so you are NOT responsible for any file input or console output. Please don’t modify the codes for I/O part. Algorithm 1 A* Algorithm 1: Put the start state S on the priority queue, called OPEN 2: while OPEN is not empty do 3: Remove from OPEN and place in CLOSED a state n for which f(n) is the minimum 4: if n is a goal state then 5: exit(trace back pointers from n to S) 6: end if 7: Expand n, generating all its successors and attach to them pointers back to n. 8: for each successor n 0 of n do 9: if n 0 is not in OPEN or CLOSED then 10: Estimate h(n 0 ), g(n 0 ) = g(n) + c(n, n0 ), f(n 0 ) = g(n 0 ) + h(n 0 ) 11: Place n 0 on OPEN. 12: else 13: if g(n 0 ) strictly less than its old g value in OPEN or CLOSED then 14: Update pointers backward from n 0 to n 15: if n 0 is in OPEN then 16: Update g(n 0 ) in OPEN. 17: else 18: Remove n 0 from CLOSED and place it on OPEN with new g(n 0 ) 19: end if 20: end if 21: end if 22: end for 23: end while 24: Exit with failure. Heuristic function For this problem, you are required to try different heuristic functions to see their effects on A* algorithm. The first heuristic function is h1(s) = 8 − max(n1, n2, n3), (1) where n1, n2, n3 are number of blocks in the central square. This formula is obtained by relaxing the constraints and allowing the blocks to switch remotely. The second one is h2(s) = 0 (2) As a result, the A* algorithm with this heuristic will degenerate into BFS. The last one should be designed by yourself. You should write down your heuristic function h3 and verify its correctness in WrittenPart.pdf. We are happy to see if you can raise a better heuristic function, but since it is an open question, so we do not require your heuristic function to beat h1. As long as you can show the validity of your heuristics and incorporate it into the code sketch correctly, then you will get full scores for this part. CS 540-1 Fall 2014 How to test We will test your program on multiple test cases, and the format of testing commands will be like this: java HW5 where inputFielName is the name of the input file which contains the initial state of the rotation game, and modeFlag is an integer ranging from 1 to 3, controlling which heuristic function your programm should use. Your program should read the initial state from the input file and then output three results sequentially: • the goal state • the operation sequence • the number of popped out states from the queue For all these three outputs, you results must match exactly to the standard ones for modeFlag=1 or 2. In cases of you may implement differently from Alg.1, we are providing you with three sample input files and six corresponding output files for modeFlag=1 or 2. They are input0.txt to input2.txt, and output01.txt to output22.txt, as seen in the zip file. For modeFlag=3, however, we will only check the validity of your result. That means we will judge whether the output goal state can be generated under your operation sequence, and whether the number of moves is same as the one computed with given heuristic. So an example command for testing for heuristic function h1 could be: java HW5 input1.txt 1 As part of our testing process, we will unzip the file you return to us, remove any class files, call javac *.java to compile your code, and call the main method of HW5 with certain parameters. Make sure that your code runs on one of the computers in the department because we will conduct our test on such computers. Deliverables 1. Hand in (see the cover page for details) your modified version of the code skeleton we provide you with your implementation of the A* search algorithm. Also include any additional java class files needed to run the program. 2. Optionally, in the written part of your homework, add any comments about the program that you would like us to know.
