Machine Problem 4: The Maze (Part II)

Introduction

You will first implement an advance (Adv) routine to move the mouse in the maze by a single position. The advance routine checks that the move is valid before moving the mouse. A valid move is one that does not put the mouse into a wall. This routine is used by both by manual mode and turbo mode. Turbo mode allows the mouse to automatically advance forward through straight hallways.

Next, you will implement the 'Auto-Advance' function which automatically forwards the mouse through hallways and around turns. Recall that at each location in the maze is marked with a decision point where two or more hallways meet. Your 'Auto-Advance' routine will advance the mouse until reaching such a decision point.

Lastly, you will implment 'Auto-Solve' which implements a Depth First Search (DFS) algorithm. This procedure will automatically guide the mouse out of the maze.

Problem Description

There are four modes of operation to this program: Manual, Turbo, AutoAdvance, and AutoSolve. Modes can be selected by pressing the 'M' key, followed by a number (0 to 3). In manual mode (M0), pressing the arrows keys will advance you by one position. In turbo mode (M1), the mouse will run down a hallway until hitting a wall. In AutoAdvance mode (M2), the mouse will traverse down a hallway until hitting a decision point. In AutoSolve mode (M3), the mouse will find the solution to the maze automatically. The mode of the program is displayed at the top-center of the screen.

In modes 1-3, the speed of the mouse can be adjusted. The (+) and (-) keys determine the delay rate. A larger rate slows the mouse, while a smaller makes the mouse move faster. The delay rate is displayed at the top-right of the screen.

Recall that the _MAZE variable contains an array of HALL, WALL, DECPOS, and ENDPOS constants. A decision point is a location in the maze where the mouse must decide which way to travel (i.e., the hallway branches). For this machine problem, we will sub-divide the decision points into three types of nodes: the original (unvisted) decision point (DECPOS), a visited decision point (VISPOS), and finished decision point (FINPOS). You can toggle the type of decision point by pressing the 'v' key.

The ASM procedures that you will need to implement are described below. There are working library versions of each of these procedures in libmp3.lib. Unless otherwise noted, it is expected that that all subroutines will preserve the value of any register that they modify. The only exception is for registers that are used as outputs of procedures.

Adv

• Purpose: Advance position by one character (if possible)
• Variables: _MAZE read, _POS modified.
• Inputs:
  • AL indicates direction (NORTH, EASH, SOUTH, WEST)
• Outputs:
  • AL: unchanged (still holds current direction)
  • AH: AH=0 indicates a valid move, AH=1 indicates invalid move
  • SI: Original position of mouse
  • DI: New position of mouse (after advancing)
• Description:

  This routine allows the mouse to move. This procedure is called with AL holding the desired direction (NORTH, EAST, SOUTH, or WEST). A move is only possible as long as the new position in the maze is NOT inside a wall. It is valid if it is the endpoint.

  The current position of the mouse in the maze is determined by the variable _POS. This value should always be returned in SI.

  If a move is possible, the new position should be computed then stored in _POS and returned in DI. Further, AH should be set to zero to indicate a valid move.

  If a move is not possible, AH should be set to 1 and DI should return the original position. (i.e., DI=SI=_POS).

AutoAdv

• Purpose: Advance the mouse through the maze until hitting a decision point.
• Inputs: AL holds starting direction
• Outputs:
  • DI: New position of mouse [like it was for Adv]
  • SI: Original position of mouse [previous value of DI]
  • AL: Final direction
• Variables: _MAZE read, _POS modified.
• Useful Constants: NORTH, EAST, SOUTH, WEST, DECPOS, VISPOS, FINPOS, ENDPOS.
• Description:

  This routine allows the mouse to move through hallways and turn at corners of the maze. The mouse will continue to explore the maze until it reaches an intersection (decision point) or hits the end of the maze.

  This routine should use Adv, UpdateScreen and Delay to move though straight hallways. When the mouse runs into a wall, it will turn and walk down the hallway further. If the mouse reaches a dead-end, it turns around and returns to the decision point where it started from.

  Decision points may be colored by the AutoSolve function. This function should stop at ANY decision point (DECPOS, VISPOS, and FINPOS), or at the end of the maze (ENDPOS). Register AL should return with the direction of the mouse when it stopped.

AutoSolve

• Purpose: Automatically solve the maze
• Inputs: AL holds starting direction
• Outputs:
  • DI: New position of mouse [like it was for Adv]
  • SI: Original position of mouse [previous value of DI]
  • AL: Final direction
• Variables: _MAZE read (decision points can be modified), _POS modified.
• Description

  This is a recursive procedure that calls itself, AutoAdv, Adv, UpdateScreen, and Delay to automatically solve the maze.

  The mouse uses a depth-first-search to find it's way to the endpoint of the maze. From a given starting point and direction, the mouse will begin by walking deeper and deeper into the maze. When the mouse hits a dead-end, it backs up, and explores another direction. In the library function, the stuck mouse next tries walking to the right. After fully exploring that region it next tries walking to the left. If, exploring all possible directions does not lead to the solution, the the mouse backs up the the previous decision point.

  The depth-first-search works by labeling decision points as unvisited, visited, or finished. Initially all decisions points are unvisited. When the algorithm reaches another unvisited decision point, it marks it as visited and recursively walks deeper into the maze. If it is found that no further progress can be made by walking in any direction (i.e., all hallways lead to a dead-end or to an already-explored region of the maze), the decision point is labeled as finished and the algorithm back-tracks to the previous decision point. To avoid infinite looping, previously-visited decision points (either visited or finished) are never re-explored. A node is only revisited if it was necessary to back-track.

  For our problem, decision points are colored blue, green, or red. Constants are defined in maze.def to define the values in _MAZE and the attribute:character value for the video display. They three types of decision points are summarized below:

  Decision-Point Type	Color	_MAZE value	Attribute: Character
  Unvisited	+ Blue	DECPOS	DECCH
  Visited	+ Green	VISPOS	VISCH
  Finished	+ Red	FINPOS	FINCH

Points

• Adv: 15 pts
• AutoAdv: 15 pts
• AutoSolve: 20 pts

Starting Files

• This program uses the same code and library routines as MP3. No additional downloads are required.
• To begin this program, just continue editing your MP3.ASM file,
• Change the comments at the top of MP3.ASM to indicate that you are working on MP4, as shown below

Relevant portions of MP3 Code for MP4