PythonBindings

MRPT Python bindings

Note

This is still a work in progress! Many classes are missing, but there are already enough to set up RBPF SLAM. Additionally, there are already ROS extensions included, which makes the usage of the bindings in combination with ROS really easy and powerful. New classes will be added step by step.

Installation

From source

For now you have to build the bindings with MRPT sources. Requirements are found and those missing are listed at he summary shown after running cmake.

Requirements

The python bindings are built using boost::python. So boost is a necessary dependency. And of course pythonlibs itself:

sudo apt-get install libboost-python-dev libpython-dev python-numpy

from apt

It is planned to release the bindings as Ubuntu (apt-deb) package.

Since 2018, pymrpt is released within the ROS package mrpt1 or mrpt2:

sudo apt-get install ros-${ROS_DISTRO}-mrpt1

Usage

For the usage of the classes itself have a look at the MRPT C++ class documentation. For the already binded classes most of the common functions are included. Some of them have been re-wrapped to have more pythonic function callings and return types, for example: methods which had &out_ variables, now usually return this, or a tuple of those, if there are more than one.

Import bindings

Supported import statements:

import pymrpt
from pymrpt import obs
from pymrpt.obs import *
from pymrpt.obs import CObservation2DRangeScan

Smart Pointers

The mrpt smart pointers are wrapped into << class_name >>Ptr-classes. They are created similar to MRPT C++ types with the static .Create() member function of their containing object class. To get acces from the pointer class to the object use .ctx() member function.

from pymrpt.obs import CObservation2DRangeScan

# create observation
obs = CObservation2DRangeScan()
# create smart pointer
obs_ptr = CObservation2DRangeScan.Create()
# set observation as new content of smart pointer
obs_ptr.ctx(obs)
# get observation object from smart pointer
obs = obs_ptr.ctx()

ROS Extensions

Several CObservation-derived classes and COccupancyGridMap2D class have .to_ROS_DataType_msg() and .from_ROS_DataType_msg() which allow easy conversion from MRPT to ROS datatypes.

# ROS imports
import rospy
from sensor_msgs.msg import LaserScan
# MRPT imports
from pymrpt.obs import CObservation2DRangeScan
from pymrpt.poses import CPose3D

# the sensor pose needed for conversion of CObservation-derived classes
# Note: Usually this is received via a call to ROS tf and a following conversion.
sensor_pose = CPose3D()

def scan_cb(scan):
    global sensor_pose
    # create mrpt observation
    obs = CObservation2DRangeScan()
    # convert ROS LaserScan to MRPT CObservation2DRangeScan
    obs.from_ROS_LaserScan_msg(scan, sensor_pose)
    # do something ...
    print 'converted ROS LaserScan to MRPT CObservation2DRangeScan'

rospy.init_node('test_conversion')
rospy.Subscriber('/scan', LaserScan, scan_cb, queue_size=10)
rospy.spin() 

Full examples

rbpf_slam.py

As the original MRPT C++ example this python script/program accepts a config file and a rawlog file to get an offline rbpf-slam running. Parameters are described by the program itself, use ./rbpf_slam.py -h to show them.

Possible dependency issues

If when applying any of the Python samples you receive the import error "ImportError: No module named pymrpt" the libraries have not been properly added to the path.

If you have installed MRPT into the usual path /usr/local, try reloading the cache with sudo ldconfig.

If you have not installed MRPT, try manually setting the variables LD_LIBRARY_PATH and/or PYTHONPATH in your shell.