Example that shows how to create a ROS node able to control the Pioneer mobile robot by IBVS visual servoing with respect to a blob. The current visual features that are used are s = (x, log(Z/Z*)). The desired one are s* = (x*, 0), with:

The degrees of freedom that are controlled are (vx, wz), where wz is the rotational velocity and vx the translational velocity of the mobile platform at point M located at the middle between the two wheels.

The feature x allows to control wy, while log(Z/Z*) allows to control vz. The value of x is measured thanks to a blob tracker. The value of Z is estimated from the surface of the blob that is proportional to the depth Z.

#include <iostream>
#include <visp/vpCameraParameters.h>
#include <visp/vpDisplayX.h>
#include <visp/vpDot2.h>
#include <visp/vpFeatureBuilder.h>
#include <visp/vpFeatureDepth.h>
#include <visp/vpFeaturePoint.h>
#include <visp/vpHomogeneousMatrix.h>
#include <visp/vpImage.h>
#include <visp/vpImageConvert.h>
#include <visp/vpServo.h>
#include <visp/vpVelocityTwistMatrix.h>
#if defined(VISP_HAVE_DC1394_2) && defined(VISP_HAVE_X11)
int main(int argc, char **argv)
try {
vpImage<unsigned char> I; // Create a gray level image container
double depth = 1.;
double lambda = 0.6;
double coef = 1./6.77; // Scale parameter used to estimate the depth Z of the blob from its surface
robot.init(argc, argv);
// Wait 3 sec to be sure that the low level Aria thread used to control
// the robot is started. Without this delay we experienced a delay (arround 2.2 sec)
// between the velocity send to the robot and the velocity that is really applied
// to the wheels.
std::cout << "Robot connected" << std::endl;
// Camera parameters. In this experiment we don't need a precise calibration of the camera
vpCameraParameters cam;
// Create a grabber based on libdc1394-2.x third party lib (for firewire cameras under Linux)
g.open(argc, argv);
// Get camera parameters from /camera/camera_info topic
if (g.getCameraInfo(cam) == false)
cam.initPersProjWithoutDistortion(600,600,I.getWidth()/2, I.getHeight()/2);
// Create an image viewer
vpDisplayX d(I, 10, 10, "Current frame");
// Create a blob tracker
vpDot2 dot;
dot.setEllipsoidShapePrecision(0.); // to track a blob without any constraint on the shape
dot.setGrayLevelPrecision(0.9); // to set the blob gray level bounds for binarisation
dot.setEllipsoidBadPointsPercentage(0.5); // to be accept 50% of bad inner and outside points with bad gray level
vpServo task;
task.setServo(vpServo::EYEINHAND_L_cVe_eJe) ;
task.setInteractionMatrixType(vpServo::DESIRED, vpServo::PSEUDO_INVERSE) ;
task.setLambda(lambda) ;
vpVelocityTwistMatrix cVe ;
cVe = robot.get_cVe() ;
task.set_cVe(cVe) ;
std::cout << "cVe: \n" << cVe << std::endl;
vpMatrix eJe;
robot.get_eJe(eJe) ;
task.set_eJe(eJe) ;
std::cout << "eJe: \n" << eJe << std::endl;
// Current and desired visual feature associated to the x coordinate of the point
vpFeaturePoint s_x, s_xd;
// Create the current x visual feature
vpFeatureBuilder::create(s_x, cam, dot);
// Create the desired x* visual feature
s_xd.buildFrom(0, 0, depth);
// Add the feature
task.addFeature(s_x, s_xd) ;
// Create the current log(Z/Z*) visual feature
vpFeatureDepth s_Z, s_Zd;
// Surface of the blob estimated from the image moment m00 and converted in meters
double surface = 1./sqrt(dot.m00/(cam.get_px()*cam.get_py()));
double Z, Zd;
// Initial depth of the blob in from of the camera
Z = coef * surface ;
// Desired depth Z* of the blob. This depth is learned and equal to the initial depth
Zd = Z;
std::cout << "Z " << Z << std::endl;
s_Z.buildFrom(s_x.get_x(), s_x.get_y(), Z , 0); // log(Z/Z*) = 0 that's why the last parameter is 0
s_Zd.buildFrom(s_x.get_x(), s_x.get_y(), Zd , 0); // log(Z/Z*) = 0 that's why the last parameter is 0
// Add the feature
task.addFeature(s_Z, s_Zd) ;
vpColVector v; // vz, wx
// Acquire a new image
// Set the image as background of the viewer
// Does the blob tracking
// Update the current x feature
vpFeatureBuilder::create(s_x, cam, dot);
// Update log(Z/Z*) feature. Since the depth Z change, we need to update the intection matrix
surface = 1./sqrt(dot.m00/(cam.get_px()*cam.get_py()));
Z = coef * surface ;
s_Z.buildFrom(s_x.get_x(), s_x.get_y(), Z, log(Z/Zd)) ;
robot.get_cVe(cVe) ;
task.set_cVe(cVe) ;
robot.get_eJe(eJe) ;
task.set_eJe(eJe) ;
// Compute the control law. Velocities are computed in the mobile robot reference frame
v = task.computeControlLaw() ;
std::cout << "Send velocity to the pionner: " << v[0] << " m/s "
<< vpMath::deg(v[1]) << " deg/s" << std::endl;
// Send the velocity to the robot
robot.setVelocity(vpRobot::REFERENCE_FRAME, v);
// Draw a vertical line which corresponds to the desired x coordinate of the dot cog
vpDisplay::displayLine(I, 0, 320, 479, 320, vpColor::red);
// A click in the viewer to exit
if ( vpDisplay::getClick(I, false) )
std::cout << "Ending robot thread..." << std::endl;
// Kill the servo task
task.print() ;
catch(vpException e) {
std::cout << "Catch an exception: " << e << std::endl;
return 1;
int main()
std::cout << "You don't have the right 3rd party libraries to run this example..." << std::endl;

Author(s): Francois Pasteau, Fabien Spindler, Gatien Gaumerais
autogenerated on Fri Jul 3 2020 03:41:43