pnf/pnf_cooc.c File Reference

#include "pnf_cooc.h"
#include <math.h>
#include <float.h>

Include dependency graph for pnf_cooc.c:

Defines
#define	INFINITY   DBL_MAX
Functions
double	pnf_cooc_detail (double lambda_i, double lambda_r, double v_i, double v_r, double delta, double *pleft, double *pbothleft, double *pmiddle, double *pbothright, double *pright, int *left, int *bothleft, int *middle, int *bothright, int *right, int *boundguard)
	Co-occurrence probability model based on "Line World Path Planning" developed by Bjoern Jensen in 2003 for joint work with Roland Philippsen on Probabilistic Navigation Functions at the Autonomous Systems Lab, EPFL.
double	pnf_cooc (double lambda_i, double lambda_r, double v_i, double v_r, double delta)
	The same as pnf_cooc_detail(), but without exposing the internal partial probabilities.
double	pnf_cooc_test_alt (double lambda_i, double lambda_r, double v_i, double v_r, double delta, unsigned int n_v_i_steps)
	For testing an alternate form of "worst case" by maximizing the expected co-occurrence over all object speeds (from 0 to v_i).

---

Define Documentation

#define INFINITY   DBL_MAX

---

Function Documentation

double pnf_cooc	(	double	lambda_i,
		double	lambda_r,
		double	v_i,
		double	v_r,
		double	delta
	)

The same as pnf_cooc_detail(), but without exposing the internal partial probabilities.

Parameters:

	lambda_i	- IN: distance to the moving object
	lambda_r	- IN: distance to the robot
	v_i	- IN: (maximum) speed of the moving object
	v_r	- IN: (maximum) speed of the robot
	delta	- IN: step size (discrete "grid")

Returns:

A value between 0 and 1 (inclusive).

double pnf_cooc_detail	(	double	lambda_i,
		double	lambda_r,
		double	v_i,
		double	v_r,
		double	delta,
		double *	pleft,
		double *	pbothleft,
		double *	pmiddle,
		double *	pbothright,
		double *	pright,
		int *	left,
		int *	bothleft,
		int *	middle,
		int *	bothright,
		int *	right,
		int *	boundguard
	)

Co-occurrence probability model based on "Line World Path Planning" developed by Bjoern Jensen in 2003 for joint work with Roland Philippsen on Probabilistic Navigation Functions at the Autonomous Systems Lab, EPFL.

Todo:

(N-1) / N can be cached. N should use ceil() instead of the flakey +0.5.

double pnf_cooc_test_alt	(	double	lambda_i,
		double	lambda_r,
		double	v_i,
		double	v_r,
		double	delta,
		unsigned int	n_v_i_steps
	)

For testing an alternate form of "worst case" by maximizing the expected co-occurrence over all object speeds (from 0 to v_i).

We do not have a closed-form solution yet, which is why this function simply tries a discrete number of speeds (could probably use Newton-Raphson, but the real deal is finding that closed-form solution anyways).

---

E* Interpolated Graph Replanner Wed Dec 12 18:55:49 2007