readme

particle filter算法代码

This is a trivial implementation of Condensation, designed for
tutorial purposes.  Simulated data is generated modelling a particle
moving in one dimension under the action of a first-order
auto-regressive process with measurements corrupted by noise. The
Condensation algorithm then tracks the particle using a model of the
same form, and the state distribution is displayed at each timestep
using an ASCII histogram, along with the estimated, true and measured
particle positions. The estimated position is displayed as an 'x', the
measured position a '+' and the true position a '.' - if any two
coincide in the same character position a '*' is displayed there
instead.

To build the program, first edit the makefile. The default compiler is 
`cc' and this can be changed if necessary. By default, the option
-DANSI_TERM_SEQUENCES is sent to the compiler, which causes the
program to generate control codes to scroll the cursor up the terminal 
after each iteration and then sleep(1), resulting in an `animated'
histogram. On a Windows machine, this option should be removed from
the CCOPTS line of the makefile to allow successful compilation. When
the makefile has been edited, type `make' and the executable
`condense' will be produced. Now run `./condense'

All of the parameters of the simulation and process models can be
changed by altering model_parameters.h and recompiling.

Files in the distribution:

README               This file.
makefile
condensation.h       Declarations of external functions and variables.
data_types.h         Declarations of data structures used.
model_parameters.h   Constants specifying the process and observation
                     models used.
utility.c            Routines to generate Gaussian noise and evaluate
                     a Gaussian density.
model_specific.c     Routines for sampling from the process model,
                     evaluating the observation density, generating
                     simulated data and displaying the estimated
                     distribution. These routines can be replaced by
                     more complex models for a realistic application.
condensation.c       The generic condensation algorithm. The routines
                     in this file implement the basic condensation
                     algorithm and can be used with any process or
                     observation model.