A Flexible and Scalable SLAM System with Full
3D Motion Estimation
Stefan Kohlbrecher and Oskar von Stryk
Technische Universit
¨
at Darmstadt
Hochschulstraße 10
Darmstadt, Germany
kohlbrecher,stryk@sim.tu-darmstadt.de
Johannes Meyer and Uwe Klingauf
Technische Universit
¨
at Darmstadt
Petersenstraße 30
Darmstadt, Germany
meyer,klingauf@fsr.tu-darmstadt.de
Abstract—For many applications in Urban Search and Rescue
(USAR) scenarios robots need to learn a map of unknown
environments. We present a system for fast online learning of
occupancy grid maps requiring low computational resources.
It combines a robust scan matching approach using a LIDAR
system with a 3D attitude estimation system based on inertial
sensing. By using a fast approximation of map gradients and
a multi-resolution grid, reliable localization and mapping ca-
pabilities in a variety of challenging environments are realized.
Multiple datasets showing the applicability in an embedded hand-
held mapping system are provided. We show that the system
is sufficiently accurate as to not require explicit loop closing
techniques in the considered scenarios. The software is available
as an open source package for ROS.
Keywords: Simultaneous Localization and Mapping, Inertial
Navigation, Robust and Fast Localization
I. INTRODUCTION
The ability to learn a model of the environment and to
localize itself is one of the most important abilities of truly
autonomous robots able to operate within real world envi-
ronments. In this paper, we present a flexible and scalable
system for solving the SLAM (Simultaneous Localization
and Mapping) problem that has successfully been used on
unmanned ground vehicles (UGV), unmanned surface vehicles
(USV) and a small indoor navigation system. The approach
consumes low computational resources and thus can be used
on low-weight, low-power and low-cost processors such as
those commonly used on small-scale autonomous systems.
Our approach uses the ROS meta operating system [1] as
middleware and is available as open source software. It honors
the API of the the ROS navigation stack and thus can easily
be interchanged with other SLAM approaches available in the
ROS ecosystem.
The system introduced in this paper aims at enabling suffi-
ciently accurate environment perception and self-localization
while keeping computational requirements low. It can be used
for SLAM in small scale scenarios where large loops do
not have to be closed and where leveraging the high update
rate of modern LIDAR systems is beneficial. Such scenarios
include the RoboCup Rescue competition, where robots have
SLAM subsystem (2D)
LIDAR
Preprocessing Scan Matching Mapping
Navigation subsystem (3D)
IMU GPS Compass
Navigation Filter
Altimeter
...
Controller
2D Pose Estimate Attitude and Initial Pose
Stabilization
Joint Values
Fig. 1. Overview of the mapping and navigation system (dashed lines depict
optional information)
to find victims in simulated earthquake scenarios which feature
rough terrain and thus require full 6DOF motion estimation of
vehicles, or the indoor navigation of agile aerial vehicles that
move fast compared to ground robots. Previous results where
the system has been used in the context of building semantic
world models in USAR environments are available in [2].
Our approach combines a 2D SLAM system based on the
integration of laser scans (LIDAR) in a planar map and an
integrated 3D navigation system based on an inertial mea-
surement unit (IMU), which incorporates the 2D information
from the SLAM subsystem as one possible source of aiding
information (Fig. I). While SLAM usually runs in soft real-
time triggered by the updates of the laser scanner device, the
full 3D navigation solution is calculated in hard real-time and
usually forms a part of the vehicle control system.
II. RELATED WORK
There has been a wealth of research into the SLAM problem
in recent years, with reliably working solutions for typical
office-like indoor scenarios using Rao-Blackwellized particle
c
2011 IEEE