Consider a simple reflex agent that is a mobile robot operating in the grid environment shown...

  

Transcribed Image Text:

Consider a simple reflex agent that is a mobile robot operating in the grid environment shown below. The robot's position can only be a grid cell and its orientation is either facing north, east, south, or west. The only possible actions are: move forward (into the next square), turn left and turn right (in place). The robot starts in a random position and orientation and its goal is to reach the position marked by an X. Every step it will perceive the environment and execute a single action based on a set of built-in condition-action rules. The robot can tell when it has reached cell X; hence the first rule is always: (RO) IF at X THEN stop Likewise all other rules are of the form: IF percept THEN action. To write them, one must consider all possible percepts available to the agent and decide the best possible action in each case. One example of a valid rule would be: (R1) IF NOT Wall_Ahead THEN Move_Forward In each of the following cases, list explicitly all percepts that are available to the robot, and provide accordingly a minimal set of rules that will achieve the goal from any given initial state, if that is possible. If it cannot be done, show your best rules and explain precisely why the task cannot be accomplished. (a) The robot is a pure reflex agent with no memory. It only has one sensor that detects if there is a wall directly ahead, or not. (1 mark) (b) The robot is like the one in part (a) except it has a second sensor that detects if there is a wall immediately to the right, or not. (2 marks) (c) The robot is like that in part (b) with the two sensors, except it now has a single-state memory, that stores percepts in the previous step. (3 marks) (d) (e) The robot is exactly like that in part (c), but now it must get out of the maze shown on the right, regardless of its starting position and orientation. (2 marks) The problem is the same as in part (d) except we remove from the maze the one thick wall (shown as 3 vertical bars). Can your rules still work? If yes, justify clearly. If not, explain the issue and how it could possibly be fixed. (2 marks) Consider a simple reflex agent that is a mobile robot operating in the grid environment shown below. The robot's position can only be a grid cell and its orientation is either facing north, east, south, or west. The only possible actions are: move forward (into the next square), turn left and turn right (in place). The robot starts in a random position and orientation and its goal is to reach the position marked by an X. Every step it will perceive the environment and execute a single action based on a set of built-in condition-action rules. The robot can tell when it has reached cell X; hence the first rule is always: (RO) IF at X THEN stop Likewise all other rules are of the form: IF percept THEN action. To write them, one must consider all possible percepts available to the agent and decide the best possible action in each case. One example of a valid rule would be: (R1) IF NOT Wall_Ahead THEN Move_Forward In each of the following cases, list explicitly all percepts that are available to the robot, and provide accordingly a minimal set of rules that will achieve the goal from any given initial state, if that is possible. If it cannot be done, show your best rules and explain precisely why the task cannot be accomplished. (a) The robot is a pure reflex agent with no memory. It only has one sensor that detects if there is a wall directly ahead, or not. (1 mark) (b) The robot is like the one in part (a) except it has a second sensor that detects if there is a wall immediately to the right, or not. (2 marks) (c) The robot is like that in part (b) with the two sensors, except it now has a single-state memory, that stores percepts in the previous step. (3 marks) (d) (e) The robot is exactly like that in part (c), but now it must get out of the maze shown on the right, regardless of its starting position and orientation. (2 marks) The problem is the same as in part (d) except we remove from the maze the one thick wall (shown as 3 vertical bars). Can your rules still work? If yes, justify clearly. If not, explain the issue and how it could possibly be fixed. (2 marks)