We're having problems getting vfh to work.  I have revised the sample code vfh.cc and am trying to run it on a Pioneer.  Attached is the vfh.cc code, our config file, and an output file that shows the messages from Player (which look fine).  The problem is that the vfh.cc code hangs when the "vfh_pp.SetMotorState(1)" command is reached.  Does anyone know what the problem is?

Thanks,
--Lynne


At 09:00 AM 12/2/2003 -0800, Brian Gerkey wrote:


hi Balajee,

I'm forwarding this to our users' list, where such questions are best
directed.

No, I don't have an example of what you're looking for, i.e., using
amcl->vfh to make a goto.  However, I've done it before and it's not
that hard to do.   The client would look something like this:

  open localize device (amcl)
  open position device (vfh)

  for(;;)
  {
    read new data from both devices

    calculate coordinate transform between vfh and amcl frames (amcl
      uses a global frame, while vfh uses an odometric frame)

    if(we have a new global goal)
    {
      convert goal from amcl frame to vfh frame

      give converted goal to vfh
    }
  }


I'll also take this opportunity to announce a new related driver, called
'wavefront'.  It is a wavefront propagation path-planner written by
Andrew, which I've integrated as a Player driver.  This driver supports
the position interface, and requires an underlying localize device and
an underlying position device that is capable of doing obstacle-avoiding
position control.  It's meant to be used like so:

            <-- amcl <--  (various sensors)
  wavefront
            <--> vfh <--> (some real position device)

This driver allows for the mythical 'global goto', by computing a list of
waypoints that form a path to the goal, and handing these waypoints off in
sequence to the underlying navigation algorithm (i.e., vfh), which controls
the robot base.

It's checked into CVS and works reasonably well, both with real bots
and in Stage.  However, it's not been thoroughly tested, so beware.
In particular, current drawbacks include:

  - no way for the client to get the list of waypoints
  - no way for the client to detect a 'no feasible plan' condition
  - no automatic replanning

Also, since we're relying on amcl for the true global pose, amcl must
be reasonably well initialized before you start pursuing a goal.

Since it's fairly complicated to set up, and since there's no
documentation yet for the wavefront driver, appended below is an example
Player config file.

        brian.

-----------------------------------------------------------------------------
# Use the following two lines to connect to a real robot:
#position:2 ( driver "trogdor" port "/dev/ttyUSB1" )
#laser:0 ( driver "sicklms200" port "/dev/ttyUSB0" invert 1)

# Use the following two lines to connect to Stage:
position:2 ( driver "passthrough" port 6665)
laser:0 ( driver "passthrough" port 6665)

# Connect amcl to robot base and laser
localize:0
(
  driver "amcl"
  odom_index 2

  laser_index 0
  laser_map "gates1-crop-clean.pgm"
  laser_map_scale 0.1
  laser_map_negate 0

  init_pose [-20 9 0]
  init_pose_var [1 1 180]

  pf_max_samples 2000
  pf_min_samples 100

  update_thresh [0.1 5]

  odom_drift[0] [0.2 0.0 0.0]
  odom_drift[1] [0.0 0.2 0.0]
  odom_drift[2] [0.2 0.0 0.2]

  enable_gui 1
  alwayson 1
)

# Connect vfh to robot base, and set various thresholds
position:1
(
  driver "vfh"
  position_index 2
  laser_index 0
  cell_size 0.1
  window_diameter 61
  sector_angle 1
  robot_radius 0.225
  safety_dist 0.15
  max_speed 0.50
  max_turnrate 75
  free_space_cutoff  1000000.0
  weight_desired_dir 5.0
  weight_current_dir 3.0
  distance_epsilon 0.5
  angle_epsilon 10
)

# Connect wavefront to vfh and amcl, and set various thresholds.  Note that
# these thresholds should be slighly looser than those used by vfh.
position:0
(
  driver "wavefront"
  position_index 1
  localize_index 0
  map_filename "gates1-crop-clean.pgm"
  map_scale 0.1
  #cspace_filename "gates1-crop-clean.cspace"
  robot_radius 0.225
  safety_dist 0.25
  max_radius 1.0
  dist_penalty 1.0
  distance_epsilon 1.0
  angle_epsilon 15
  #alwayson 1
)

-----------------------------------------------------
Prof. Lynne E. Parker                               
Department of Computer Science      
The University of Tennessee             
URL: http://www.cs.utk.edu/~parker

Non scholae, sed vitae discimus.
------------------------------------------------------