Differences

This shows you the differences between two versions of the page.

--- public:t-622-arti-15-1:prog1 [2015/01/22 17:05] – stephan
+++ public:t-622-arti-15-1:prog1 [2024/04/29 13:33] (current) – external edit 127.0.0.1
@@ Line 23: / Line 23: @@
 Your actions have the following costs:
-  * 1 + 15*D, if you TURN_OFF the robot in the home location and there are D dirty cells left
+  * 1 + 50*D, if you TURN_OFF the robot in the home location and there are D dirty cells left
-  * 100 + 15*D, if you TURN_OFF the robot, but not in the home location and there are D dirty cells left
+  * 100 + 50*D, if you TURN_OFF the robot, but not in the home location and there are D dirty cells left
   * 5 for SUCK, if the current location of the robot does not contain dirt
   * 1 for SUCK, if the current location of the robot contains dirt
@@ Line 46: / Line 46: @@
   - Post the results (number of state expansions, maximum size of the frontier, cost of the found solution) on Piazza, to see how well you are doing compared to the other students.
   - Optionally (up to 10% bonus points): Implement detection of revisited states in A*-Search. Re-run the experiments and comment on the difference in the results.
-  - Try to improve your heuristics, but make sure that it does not increase the overall runtime of search algorithm. You can easily create a heuristics that tells you the optimal costs but is as expensive as a blind search. Try to avoid this.
+  - Try to improve your heuristics while keeping it admissible, but make sure that it does not increase the overall runtime of the search algorithm. You can easily create a heuristics that tells you the optimal costs but is as expensive as a blind search. Try to avoid this.
 ===== Material =====
@@ Line 73: / Line 73: @@
   * Implement a class "State" that contains all the information about a state that you need to keep track of. The State class should have a method that returns a list of all moves that are legal in the state and a method that takes a move and returns the state that results from executing the move.
   * Distinguish between nodes of the search tree and states! Each node has an associated state, but there may be several nodes that have the same state associated. In addition to a state, a node contains a reference to its parent (unless it is the root node), the move that was taken to get to the node and (if necessary) the path cost associated for getting from the root node to the node.
+  * Keep the code for all the algorithms and make it easy to switch between them (e.g., with a command line parameter or by changing a single line of code).
 For developing an admissible heuristics:
@@ Line 92: / Line 93: @@
 ===== Grading =====
-* 70% - implementation (correct implementation of the model, algorithms and heuristics, quality/readability of the code)
+  * 70% - implementation (correct implementation of the model, algorithms and heuristics, quality/readability of the code)
-* 30% - report (answers to questions and report on results)
+  * 30% - report (answers to questions and report on results)
-Bonus points (up to 5%) for the teams that find the optimal solution for most of the environments and/or have the fastest algorithms and another 10% for (correctly) implementing detection of revisited states and reporting on those results.
+Bonus points:
+  * up to 5% for the agents that find the optimal solution for the most environments fastest
+  * up to 10% for (correctly) implementing detection of revisited states and reporting on those results (task 9)