AStarGazer
==========

AStarGazer is a program for visualizing the A* search algorithm on a tile map.

The A* search algorithm (en.wikipedia.org/wiki/A_Star_Search_Algorithm) is 
used to traverse a connection of nodes, called a graph. This implementation 
uses it to find a path between a starting point and a goal point on a two-
dimensional grid, a problem often encountered in tile-based video game AI. It 
also displays the inner workings of the algorithm by coloring the open and 
closed sets of nodes on the map and drawing a line representing the current 
path being investigated.

Open set - The collection of nodes that have yet to be investigated

Closed set - The collection of nodes that have already been investigated

Heuristic - The method for calculating the cost of a node

Neighbor Selection - The method for determining which nodes can be reached from 
the current node

With each step, neighboring nodes are selected for analysis. If a neighbor is 
considered a valid move, its cost is calculated. The cost (f) is the sum of two 
parts: 

1) the cost already spent to get to the current node (g), which is the current 
location's cost plus whatever incremental cost is required to get to the 
neighboring node; and
2) the estimated cost (h) to get to the goal.

All these neighbors are added to the open set, which is the collection of nodes 
to be investigated, and the current location is added to the closed set, as it 
has already been analyzed. From all the points stored on the open set, the one 
with the lowest cost is investigated next.  The A* search algorithm can be used 
to search through graphs that are not representations of physical space. A 
common example is the sliding number puzzle where 8 or 15 numbers are stored in 
a square grid. The numbers can be moved using the single empty spot, and each 
arrangement of puzzle pieces can be considered a node in the graph with an 
associated heuristic of the number of moves taken so far added to the estimated 
distance the numbers are from their ordered locations. 

Usage Instructions:

Step - Moves the algorithm forward one step. A step is defined as popping the 
       minimum cost node from the open set, analyzing its cost with a 
       heuristic, and adding its valid neighbors to the open set.

Solve - Continues to step until the algorithm is complete.

Reset - Resets the algorithm to its initial state to be re-run.

Generate - Generates a new random map and start and goal points.

Heuristics - Select which heuristic to use, meaning how the distance between 
             two points should be evaluated in the algorithm.

    Manhattan - Manhattan (or Taxicab) distance only permits traveling along 
                the x and y axes, so the distance between (0, 0) and (1, 1) is 
                2 from traveling along the X axis 1 unit and then up the Y 
                axis 1 unit, and not the hypotenuse, which would be sqrt(2)/2. 
                It's so named because in a city you're bound to the streets. 
                You can't cut diagonally through a building to go north/south 
                one block and east/west one block.
    Chebyshev - Chebyshev distance considers the movement on both axes
                and takes the maximum between the two differences. This 
                heuristic can be paired with the 8-directional neighbor. 
                To visualize the usefulness of this heuristic imagine a
                claw machine. While you're moving in one direction you can  
                simultaneously move in the other direction at no cost in time 
                (unless you move more in this direction, which makes this the max).
                Sometimes in poorly programmed video games you notice that your 
                character can move faster on the diagonal. This is because the 
                motion is working in Chebyshev-space, but the world map you're 
                running around on is in Euclidean space. Say the character has 
                a set speed of one unit per update cycle. The up and the left 
                buttons are pressed on the controller simultaneously, so the 
                character moves up one unit and left one unit. The effect on 
                the screen in Euclidean space is that the character gets a 
                speed boost. The character should have been limited to one unit 
                of motion, but by moving in two dimensions at once, the 
                character traveled sqrt(2)/2 units in one update. In Chebyshev 
                space these movements are considered equal, and the discrepancy 
                between the Euclidean space of the game world and the Chebyshev 
                space of the poorly chosen input scheme betray this. 
    Diagonal -  Diagonal distance is a perfect pair with the 8-directional 
                neighbor selection scheme. It performs a perfect measure in  
                distance when movement is restricted entirely to horizontal,  
                vertical, and diagonal motion.
    Euclidean - This is normal distance calculated by the Pythagorean formula: 
                d = sqrt((a_x-b_x)^2+(a_y-b_y)^2)
    Euclidean Squared - This heuristic is an optimized version of the 
                Euclidean heuristic. It is the normal distance formula, but 
                without taking the square root. Because the distance is only 
                used to compare, comparing the squares of two values still 
                works to determine which path is shorter. Is does weight the 
                from-cost (g) much less than the to-cost (h) because the from-
                cost is added up incrementally while the to-cost is calculated 
                in one big piece. This is caused because the distance between 
                the start and the cursor is many very small steps squared 
                individually before being added up (often just 1, which is not 
                increased by being squared at all) so the Euclidean squared 
                heuristic is much greedier about getting to its goal. 
                Sometimes this imbalance will cause it to select a less 
                optimal path closer to the start in favor of a better path 
                closer to the goal.

Neighbors - Select the method by which neighboring tiles are selected.

    4-directional - Neighbors are selected from the traversable tiles to the 
                    north, east, south, and west only. This works best with 
                    the Manhattan heuristic.
    8-directional - Neighbors are selected from diagonally adjacent tiles in 
                    addition to the north, east, south, and west directions. 
                    This works best with the Euclidean heuristic. Combining 
                    this neighbor selection scheme with the Manhattan 
                    heuristic can yield strange results because a diagonal 
                    path over a single tile is counted as a distance of 2.
    Jump Point -    Neighbors are selected by extending rays out from the 
                    current tile until an obstacle or the goal is hit. These 
                    tiles are "neighbors" in the sense that they are to be 
                    used in the next step of the algorithm, but are not 
                    necessarily adjacent.

Colors - Select a color scheme for the map panel.

Obstacles - Select the type of obstacles to place on the tile map.

    Blocks - Rectangular blocks will be randomly placed in the map.

    Rounded Edges - The same rectangular blocks will be placed in the map, 
                    but their edges will be rounded off for an organic feel.

    Lines - A series of vertical or horizontal lines will be placed in the map 
                    with a chance of one or two perpendicular lines as well.

    Random Points - Many single randomly selected tiles will be marked as 
                    non-traversable in the map.

    Maze - A maze will be generated using the recursive backtracker maze 
                    generation method.

Solve Delay - How long the solve process wait before running the next step.

Zoom - Zoom in and out of the map panel.

Full Dijkstra Search (h=0) - This option sets the h value to zero, meaning 
                     there is no incentive for the algorithm to select a good 
                     path to test first. It treats all points equally in terms 
                     of how far is left to travel. This converts the A* search 
                     algorithm into the more generalized Dijkstra's algorithm.

Randomize Equicost Nodes - Selecting this option will shuffle the neighboring 
                     points before they are sorted prior to being pushed onto 
                     the open set. One feature of the Manhattan heuristic is 
                     that on a full grid it doesn't matter when you choose to 
                     go along the X axis and when you choose to go along the Y 
                     axis. You could travel only in the Y direction until you 
                     were at the correct row, and then travel only in the X 
                     direction until you reach your destination. Alternatively 
                     you could zigzag with every step. The distance traveled 
                     is the same in both cases. This means that there are lots 
                     of equidistant paths to decide between when using a 
                     Manhattan heuristic. The decision to try one direction 
                     instead of another frequently becomes a factor of which 
                     is added to the set first, and the sort will just leave 
                     equal cost nodes in their original ordering. Randomizing 
                     them will prevent the heuristic from favoring one 
                     direction simply because tiles in that direction are 
                     added to the set before tiles in another direction.

Show Grid - This option determines whether to display the tile grid on the map

Seed Map Generator - Selecting this option from the File menu will permit you 
                     to specify a seed value for the random number generator 
                     that builds the tile maps. You can enter an integer to use 
                     or a unique integer can be generated using text you enter. 
                     Seeds are your keys to revisiting maps as each key will 
                     regenerate the same unique map each time it is used as the 
                     seed.

Seed values for notable maps (for Block type obstacles):

Randomizing equicost nodes can dramatically affect the route in these maps:
1260, 1275

Maps that take the algorithm a lot of computation to traverse:
485, 759, 769, 1178, 1271, 2264, 733, 1124, 1616, 3850

These maps show off the distinction between the Euclidean and the Euclidean 
squared heuristics well:
264, 393, 747, 1115, 7478, 3902

There is a race effect between multiple routes when using some heuristics in 
these maps, especially with the full Dijkstra search enabled:
534, 804, 1503, 1626, 1881, 2509, 1069, 1305, 2241, 2986

These maps feature drastic shortcuts when a neighbor selection scheme that 
permits diagonal traversal, like the 8-directional scheme, is used:
824, 1142, 1531, 2013, 2065, 2168, 2376, 2546, 3019, 3161, 3657, 3819, 3911, 
6206

Finding a route in these maps is only possible for a neighbor selection scheme 
that permits diagonal traversal, like the 8-directional scheme:
527, 796, 1010, 1277, 1304, 1355, 1384, 1737, 1868, 2000, 2142, 2272, 2346, 
2430, 2602, 2677, 2704, 2812, 3012, 3030, 3287, 3468, 3763, 3898

Trivial maps where the start and goal are on the same point:
2203, 2517

These maps aren't especially unique, but feature interesting solutions:
 172,  185,  235,  259,  364,  378,  388,  402,  466,  468,  484,  596,  674, 
 694,  709,  713,  719,  742,  756,  760,  763,  782,  791,  814,  858,  863, 
 885,  937, 1163, 1234, 1377, 1403, 1469, 1570, 1597, 1614, 1619, 1623, 1628, 
1629, 1633, 1645, 1650, 1678, 1686, 1705, 1786, 1788, 1810, 1812, 1842, 1983, 
2006, 2038, 2145, 2172, 2219, 2234, 2243, 2256, 2292, 2308, 2317, 2348, 2367, 
2434, 2439, 2558, 2559, 2588, 2639, 2673, 2707, 2711, 2718, 2725, 2770, 2773, 
2782, 2841, 2904, 2930, 2934, 2962, 2963, 2973, 3051, 3110, 3171, 3270, 3278, 
3348, 3417, 3451, 3462, 3479, 3481, 3489, 3507, 3518, 3542, 3551, 3658, 3670, 
3686, 3690, 3711, 3736, 3758, 3767, 3892, 3907, 3912, 3944, 3952, 3980, 3982, 
3988, 4004