Multi-robot Environment Exploration
Based on Label Maps Building
via Recognition of Frontiers
Bin Wang
School of Automation Science and Electrical Engineering
Beihang University
Beijing, China
physicsexman@gmail.com
Shiyin Qin
School of Automation Science and Electrical Engineering
Beihang Unviersity
Beijing, China
qsy@buaa.edu.cn
Abstract—This paper presents a new approach to multi-robot
environment exploration based on label maps building through
recognition of frontiers. At first, the model of multi-robot
environment exploration is built and analysed, in which, the label
map building, the formation and role modeling, and the task
assignment are synthetically considered. Then the behavior
coordination towards exploration process is studied in depth
according to label maps and cooperative tasks. Thus the
extraction of shape context features from local maps, label map
building based on frontier recognition with shape context,
behavior coordination with label maps and cooperative tasks, and
real time path planning and updating are integrated to make up
of an effective implementing algorithm which can successfully
avoid some repeated and/or redundant exploration of the same
region by multiple robots. A series of simulations and real world
experiments results demonstrate the performance advantages of
our proposed approach so as to outperform the other methods
ignoring label maps in the same conditions.
Keyword—Multi-robot; unknown environment exploration;
behavior coordination; task assignment; shape context; label maps
I. INTRODUCTION
Exploration in the unknown environment by mobile robots
system is necessary in some real world situation nowadays [1].
Mobile robot exploration has been used in lots of occasions
such as rescue mission, space and planetary exploration, floor-
cleaning and so on [2]. Exploration task aims at gathering the
sensor information of the environment to detect all the
unknown place and to build a map. Research has gathered great
improvement in the last forty years, especially for the space
exploration. However, to continue gaining new knowledge
requires new capability. Autonomous designed for the
environment exploration should have the abilities of navigation
on different kind of surface (rocks, ruins, etc.). Additional,
sensing and perception for exploration missions as well as
landmark mapping are problems to be addressed on the space
exploration problems [3]. In recent years, coordination and
cooperation have been considered in the exploration problem.
Multi-robot system has lots of advantages than single robot.
First of all, multi robots have the ability of redundancy, they
can be more fault-tolerant than single robot. Secondly, one
robot costs more time than multi-robot to explore unknown
place. Last but not the least, more than one robot explore
together can merge overlapping information to reduce
uncertainty. Although coordination exploration provides
exploring tasks more quick and robust qualities, a good
algorithm is required for multi-robot coordination and
cooperation. Efficient algorithm can make the exploration
process be more quickly reliable and robust. Coordination
robots system should prevents exploring of the same area for
more than one robot, so, the exploration team should keep
cooperation to minimize the explore time and reduce potential
overlap [4].
Here, we are considering robots exploration missions in an
indoor unknown area with using the map information for the
coordination exploration. The algorithm takes the information
of the explored map into account to speed up the coordination
exploration. We extract the modified shape context information
in the explored map to recognize the corridors. A label map
with the corridor information is built. The robots system
coordinately explores the map with the label maps to get a more
uniformed distribution in the indoor environment. Though this
way, coordination robots can be distributed more uniformly to
explore different places, the detect time and overlap for robots
can be reduced as a result. Our algorithm has been realized in
the rooms in robot operation system (ROS) [5] and real-world
performance compared with the algorithm that does not think
over the utility of the labels.
II. RELATED
WORT
Multi-robot system has been developed for more than forty
years. From early stage there are many researchers used multi-
robot system to explore the environment [6-7]. Researchers
focused on the location problems and the coordination
problems most. In recent years, multi-robot system has
achieved great progress. Here, we present several state-of-art
multi-robot exploration researches in this section.
A. Task cooperation and behavior coordination towards
multi-robot environment exploration
Given the explored model of the map to choose next goal,
exploration problem is classically referred to as next best view
(NBV) problem. It means that choosing the best target of the
current set of the candidates’ goals which has the highest
expected values [8], then the robot can plan the path for the goal
and executed the arrangement. For the goals the robots should
pursued, Yamauchi proposed a very popular frontier-based
technology [9]. He presented the notion of frontiers which
represented the regions on the boundary between open space
This work is partly supported by the National Natural Science Foundation
of China (No. 60875072, 61273350) and Beijing Natural Science Foundation
(No. 4112035).