def execute(self, userdata):
        rospy.loginfo('Executing state PoseEstimateObject')

        userdata.sec_try = False
        userdata.fitted_object = None

        # Get the ID of the classified object from the YAML file
        ids = utils.get_yaml_objects_nrs(userdata.yaml, userdata.focused_object.object.id)

        # pose_estimated = perception.get_gripper_perception(pose_estimation=True, object_ids=ids)[0] # OLD
        pose_estimated_objects = perception.get_gripper_perception(pose_estimation=True, cuboid=False, object_ids=ids)
        rospy.logdebug("Returned pose_estimated_objects:")
        rospy.logdebug(str(pose_estimated_objects))

        if pose_estimated_objects is None:
            rospy.logwarn('Couldn\'t get gripper perception.')
            if not userdata.sec_try_done:
                userdata.sec_try = True
            return 'fail'

        if not pose_estimated_objects:
            rospy.logdebug('No object found for pose estimation.')
            if not userdata.sec_try_done:
                userdata.sec_try = True
            return 'fail'

        # Convert the poses to odom_combined
        map(lambda obj: utils.euroc_object_to_odom_combined(obj), pose_estimated_objects)

        # TODO Investigate the result for the object that's closest to the original object we were interested in
        # TODO find proper threshold
        corresponding_object_idx = utils.get_nearest_object_idx(userdata.focused_object, pose_estimated_objects, 0.1)

        # # TODO if more than one object were recognized, classify again and take the closest
        # if len(pose_estimated_objects) == 1:
        #     corresponding_object_idx = 0
        # else:
        #     corresponding_object_idx = None
        #     for idx in range(0, len(pose_estimated_objects)):
        #         if pose_estimated_objects[idx].mpe_success:
        #             corresponding_object_idx = idx

        # TODO Call the perception again
        if corresponding_object_idx is None:
            rospy.logdebug('Couldn\'t find the desired object on the next lookup again')
            if not userdata.sec_try_done:
                userdata.sec_try = True
            return 'fail'

        pose_estimated = pose_estimated_objects[corresponding_object_idx]
        rospy.logdebug('Use Result from Object :%s' % str(pose_estimated))

        if pose_estimated is None or not pose_estimated.mpe_success:
            rospy.logwarn('Poseestimation failed.')
            if not userdata.sec_try_done:
                userdata.sec_try = True
            return 'fail'

        userdata.fitted_object = pose_estimated

        rospy.loginfo('Publishing objects into planning scene.')
        utils.publish_collision_objects([pose_estimated.mpe_object])

        return 'success'
Esempio n. 2
0
    def execute(self, userdata):
        rospy.loginfo('Executing state ClassifyObject')
        userdata.sec_try = False

        rospy.loginfo('Using simple perception.')
        get_gripper = perception.get_gripper_perception()
        if get_gripper is None:
            rospy.logwarn("get_gripper is None!!!?!?")
            get_gripper = []
        perceived_objects = get_valid_objects(get_gripper)
        rospy.logdebug('Found ' + str(len(get_gripper)) + ' objects, ' + str(len(perceived_objects)) + ' are valid.')
        if not perceived_objects:
            if not userdata.sec_try_done:
                userdata.sec_try = True
            return 'noObject'
            # else:
            #     return 'secTry'

        collision_objects = []
        matched_objects = []
        found_object_names = []  # map(lambda obj1: obj1.mpe_object.id, userdata.objects_found)
        for obj in userdata.objects_found:
            found_object_names.append(obj.mpe_object.id)

        for obj in perceived_objects:
            # classify object
            matched_obj = classify_object(obj)
            rospy.logdebug('Matched: \n' + str(obj) + '\n----------------------------------\nwith: ' + str(matched_obj))

            if matched_obj.c_type == EurocObject.OBSTACLE:
                if matched_obj.c_cuboid_success:
                    collision_objects.append(matched_obj.object)
                continue

            if matched_obj.c_type == EurocObject.UNKNOWN or matched_obj.c_type == EurocObject.TABLE:
                if matched_obj.c_cuboid_success:
                    collision_objects.append(matched_obj.object)
                # TODO something better than ignoring the problem
                rospy.logwarn('Unknown object perceived or table. Ignoring it for now.')
                continue

            if matched_obj.c_type == EurocObject.OBJECT:
                # check if the object was already placed
                rospy.loginfo('Matched object ' + str(matched_obj.object.id) + '.')
                if matched_obj.object.id in found_object_names:
                    rospy.loginfo('Object %s was already seen.' % matched_obj.object.id)
                else:
                    utils.euroc_object_to_odom_combined(matched_obj)
                    matched_objects.append(matched_obj)

        # publish_collision_objects(collision_objects)
        userdata.classified_objects = matched_objects

        # check if it was an object
        if matched_objects:
            return 'objectsClassified'
        else:
            if not userdata.sec_try_done:
                userdata.sec_try = True
            return 'noObject'
            # else:
            #     return 'secTry'
Esempio n. 3
0
    def execute(self, userdata):
        rospy.loginfo('Executing state PoseEstimateObject')

        userdata.sec_try = False
        userdata.fitted_object = None

        # Get the ID of the classified object from the YAML file
        ids = utils.get_yaml_objects_nrs(userdata.yaml,
                                         userdata.focused_object.object.id)

        # pose_estimated = perception.get_gripper_perception(pose_estimation=True, object_ids=ids)[0] # OLD
        pose_estimated_objects = perception.get_gripper_perception(
            pose_estimation=True, cuboid=False, object_ids=ids)
        rospy.logdebug("Returned pose_estimated_objects:")
        rospy.logdebug(str(pose_estimated_objects))

        if pose_estimated_objects is None:
            rospy.logwarn('Couldn\'t get gripper perception.')
            if not userdata.sec_try_done:
                userdata.sec_try = True
            return 'fail'

        if not pose_estimated_objects:
            rospy.logdebug('No object found for pose estimation.')
            if not userdata.sec_try_done:
                userdata.sec_try = True
            return 'fail'

        # Convert the poses to odom_combined
        map(lambda obj: utils.euroc_object_to_odom_combined(obj),
            pose_estimated_objects)

        # TODO Investigate the result for the object that's closest to the original object we were interested in
        # TODO find proper threshold
        corresponding_object_idx = utils.get_nearest_object_idx(
            userdata.focused_object, pose_estimated_objects, 0.1)

        # # TODO if more than one object were recognized, classify again and take the closest
        # if len(pose_estimated_objects) == 1:
        #     corresponding_object_idx = 0
        # else:
        #     corresponding_object_idx = None
        #     for idx in range(0, len(pose_estimated_objects)):
        #         if pose_estimated_objects[idx].mpe_success:
        #             corresponding_object_idx = idx

        # TODO Call the perception again
        if corresponding_object_idx is None:
            rospy.logdebug(
                'Couldn\'t find the desired object on the next lookup again')
            if not userdata.sec_try_done:
                userdata.sec_try = True
            return 'fail'

        pose_estimated = pose_estimated_objects[corresponding_object_idx]
        rospy.logdebug('Use Result from Object :%s' % str(pose_estimated))

        if pose_estimated is None or not pose_estimated.mpe_success:
            rospy.logwarn('Poseestimation failed.')
            if not userdata.sec_try_done:
                userdata.sec_try = True
            return 'fail'

        userdata.fitted_object = pose_estimated

        rospy.loginfo('Publishing objects into planning scene.')
        utils.publish_collision_objects([pose_estimated.mpe_object])

        return 'success'