localizer.cpp

/**
	localizer.cpp

	Purpose: implements a 2-dimensional histogram filter
	for a robot living on a colored cyclical grid by 
	correctly implementing the "initialize_beliefs", 
	"sense", and "move" functions.

	Programmer:		Daniel Rines
    Date Created:	2019.06.06
    Date Revised:	2019.06.07
*/

#include "localizer.h"
#include "helpers.cpp"
#include <stdlib.h>
#include "debugging_helpers.cpp"

using namespace std;

/**
    Initializes a grid of beliefs to a uniform distribution. 

    @param grid - a two dimensional grid map (vector of vectors 
    	   of chars) representing the robot's world. For example:
    	   
    	   g g g
    	   g r g
    	   g g g
		   
		   would be a 3x3 world where every cell is green except 
		   for the center, which is red.

    @return - a normalized two dimensional grid of floats. For 
           a 2x2 grid, for example, this would be:

           0.25 0.25
           0.25 0.25
*/
vector< vector <float> > initialize_beliefs(vector< vector <char> > grid) {
	// initialize local variables for the grid variables
	int height = grid.size();
    int width = grid[0].size();
    int area = height * width;

	// ratio factor creating the uniform distribution
    float belief_per_cell = 1.0 / area;
    
	// newGrid with uniform ratio values populating all of the newGrid cells
	vector < vector<float> > newGrid (height, vector<float>(width, belief_per_cell));
	
	// returns the uniform newGrid structure
	return newGrid;
}

/**
    Implements robot motion by updating beliefs based on the 
    intended dx and dy of the robot. 

    For example, if a localized robot with the following beliefs

    0.00  0.00  0.00
    0.00  1.00  0.00
    0.00  0.00  0.00 

    and dx and dy are both 1 and blurring is 0 (noiseless motion),
    than after calling this function the returned beliefs would be

    0.00  0.00  0.00
    0.00  0.00  0.00
    0.00  0.00  1.00 

  @param dy - the intended change in y position of the robot

  @param dx - the intended change in x position of the robot

    @param beliefs - a two dimensional grid of floats representing
         the robot's beliefs for each cell before sensing. For 
         example, a robot which has almost certainly localized 
         itself in a 2D world might have the following beliefs:

         0.01 0.98
         0.00 0.01

    @param blurring - A number representing how noisy robot motion
           is. If blurring = 0.0 then motion is noiseless.

    @return - a normalized two dimensional grid of floats 
         representing the updated beliefs for the robot. 
*/
vector< vector <float> > move(int dy, int dx, 
  vector < vector <float> > beliefs,
  float blurring) {
	
  	// initialize local variables based on the beliefs matrix dimensions
  	int height = beliefs.size();
  	int width = beliefs[0].size();

	// construct a newGrid that is the same size as the beliefs matrix
	// and fully initialize it with zeros:
	vector < vector <float> > newGrid (height, vector <float> (width, 0.0));

	// create local variables for new position iterators (new_i, new_j)
	int new_i = 0;
	int new_j = 0;

	// loop thru the beliefs matrix and shift cell values based on dy, dx
	for (int i = 0; i < height; i++) {
		for (int j = 0; j < width; j++) {

			// calculate the frame shift for newGrid
			new_i = ((i + dy) % height + height) % height;
			new_j = ((j + dx) % width + width) % width;

			// store the value from beliefs into the correct newGrid pos.
			newGrid[new_i][new_j] = beliefs[i][j];
		}
	}

	// return the newGrid matrix + passed in blurring value
	return blur(newGrid, blurring);
}


/**
    Implements robot sensing by updating beliefs based on the 
    color of a sensor measurement 

	@param color - the color the robot has sensed at its location

	@param grid - the current map of the world, stored as a grid
		   (vector of vectors of chars) where each char represents a 
		   color. For example:

		   g g g
    	   g r g
    	   g g g

   	@param beliefs - a two dimensional grid of floats representing
   		   the robot's beliefs for each cell before sensing. For 
   		   example, a robot which has almost certainly localized 
   		   itself in a 2D world might have the following beliefs:

   		   0.01 0.98
   		   0.00 0.01

    @param p_hit - the RELATIVE probability that any "sense" is 
    	   correct. The ratio of p_hit / p_miss indicates how many
    	   times MORE likely it is to have a correct "sense" than
    	   an incorrect one.

   	@param p_miss - the RELATIVE probability that any "sense" is 
    	   incorrect. The ratio of p_hit / p_miss indicates how many
    	   times MORE likely it is to have a correct "sense" than
    	   an incorrect one.

    @return - a normalized two dimensional grid of floats 
    	   representing the updated beliefs for the robot. 
*/
vector< vector <float> > sense(char color, 
	vector< vector <char> > grid, 
	vector< vector <float> > beliefs, 
	float p_hit,
	float p_miss) 
{
  	// initialize local variables based on the beliefs matrix dimensions
  	int height = beliefs.size();
  	int width = beliefs[0].size();

	// construct a newGrid that is the same size as the beliefs matrix
	// and fully initialize it with zeros:
	vector < vector <float> > newGrid (height, vector <float> (width, 0.0));

	// loop thru the grid and beliefs matrices to establish the new beliefs
	// construct some local variables for comparing the hit truth in sense 
	int hit = 0;

	// start with the first row and loop through the cols:
	for (int i = 0; i < height; i++) {
		for (int j = 0; j < width; j++) {

			// check if we have a color match (hit)
			hit = (color == grid[i][j]) ? true : false;

			// assign the newGrid (beliefs) values
			newGrid[i][j] = beliefs[i][j] * (hit * p_hit + (1 - hit) * p_miss);
		}
	}

	// first normalize the newGrid values and return new beliefs
	return normalize(newGrid);
}