model_selection_client.py

Go to the documentation of this file.

#!/usr/bin/env python

import roslib; roslib.load_manifest('door_perception')
import rospy
import numpy

from articulation_msgs.msg import *
from articulation_msgs.srv import *
from geometry_msgs.msg import Pose, Point, Quaternion
from sensor_msgs.msg import ChannelFloat32


PRISMATIC = 0
ROTATIONAL = 1
MODELS={PRISMATIC:'prismatic',ROTATIONAL:'rotational'}

def sample_track(model = PRISMATIC, n = 100, sigma_position = 0.02):
        msg = TrackMsg()
        msg.header.stamp = rospy.get_rostime()
        msg.header.frame_id = "/"
        msg.id = model
        for i in range(0,n):
                q = i / float(n)
                if model == PRISMATIC:
                        pose = Pose(Point(q, 0, 0), Quaternion(0, 0, 0, 1))
                elif model == ROTATIONAL:
                        pose = Pose(Point(numpy.sin(q), numpy.cos(q) - 1.0, 0), Quaternion(0, 0, 0, 1))
                else:
                        raise NameError("unknown model, cannot generate trajectory!")
                pose.position.x += numpy.random.rand()*sigma_position
                pose.position.y += numpy.random.rand()*sigma_position
                pose.position.z += numpy.random.rand()*sigma_position
                msg.pose.append( pose )
        return msg

def main():
        rospy.init_node('test_fitting')

        model_select = rospy.ServiceProxy('model_select', TrackModelSrv)
        model_pub = rospy.Publisher('model', ModelMsg)
        print
  
        while True:
                for model_type,model_name in MODELS.items():
                        request = TrackModelSrvRequest()
                        print "generating track of type '%s'" % model_name
                        request.model.track = sample_track( model_type )
                 
                        try:            
                                response = model_select(request)
                                print "selected model: '%s' (n = %d, log LH = %f)" % (
                                        response.model.name,
                                        len(response.model.track.pose),
                                        [entry.value for entry in response.model.params if entry.name=='loglikelihood'][0]
                                )
                                model_pub.publish(response.model)
                        except rospy.ServiceException: 
                                print "model selection failed"
                                pass
                        if rospy.is_shutdown():
                                exit(0)
                        print
                        rospy.sleep(0.5)



if __name__ == '__main__':
  main()