SURVEY AND ANALYSIS OF DIFFERENT METHODOLOGIES USED

IN MOBILE ROBOT NAVIGATION

Captain Zin Maung Cho, Captain Myo Min Tun

ABSTRACT

Robot navigation is a fundamental problem in robotics. Navigation

associated to mobile robot is the problem of finding a feasible path form one
configuration to another by avoiding the obstacles along its path. Combination
of different approaches is usually used in order to make the local as well as
global navigation system more effective. This paper provides an overview on
different methodologies used for the successful navigation of mobile robot in
last few decades. Finally this paper points out the scopes and challenges which
are still in this field.

Keywords: mobile robot, navigation, mobile robot navigation, local navigation,

global navigation.
1. INTRODUCTION
Robot navigation is a fundamental problem in robotics. Navigation related
to mobile robot is the ability of finding a collision free path from its starting
position to the goal position by avoiding the obstacles. Moreover the selected
path should be optimized (i.e. having smallest possible distance and minimum
number of turns) to make sure that least amount of energy and time are used by
the robot in roaming from starting point to its target. In order to navigate
successfully in an unknown or partially known environment without any human
assistance, the mobile robots should be able to extract the necessary information
from the environment, use their built-in knowledge for perception and to take the
action required to plan a feasible path for collision free motion and to reach the
goal see in Figure 1. A lot of research work has been carried out in order to solve
this problem.
Figure 1. Navigation Control Loop

2. Robot Navigation Techniques

Most mobile robot navigation techniques are developed at two levels:
local level and global level. Local navigation by reactive means is employed by
techniques such as fuzzy systems, or field-based systems. In this a robot relies
only on current or recent sensor data. This is useful for rapid responses to avoid
collisions. Global navigation is used to break deadlocks. Global navigation looks
at the broader objective of the robot, and identifies long-range paths for the robot
to follow. Global navigation techniques which are based on static environment
are not so applicable for mobile robot navigation.

3. Local Navigation
There are many kinds of local navigation systems. The most commonly
used methods are described here.

3.1. Artificial Potential Fields

The “Artificial Potential Fields” (APF) method involves modeling tile
robot as a particle in space, acted on by some combination of attractive and
repulsive fields [6]. In this technique obstacles and goods are modeled as
charged surfaces, and the net potential generates a force on the robot. These
forces push the robot away from the obstacles, while pulling it towards the goal.
The robot moves in the direction of greatest negative gradient in the potential.

Remark: Despite its simplicity this method can be effectively used in

many simple environments.

3.2. Vector Field Histogram (VFH) Technique

Borenstein and Koren [3] devised the “Vector Field Histogram’’ (VFH)
technique. The VFH uses local maps consisting of occupancy grids. These grids
are transformed into maps where the occupied cells contain peaks, and open
spaces contain troughs. The robot is then drawn towards the troughs. Ulrich and
Borenstein extended this into VFH+ method.

Remark: This method includes the robots kinematic constraints when

generating the troughs and peaks.

3.3. Dynamic Window Approach

The method is applied to move the robot using a model of the positions
attainable by the robot. Fox developed a “Dynamic Window Approach” (DWA)
as a function of the robot velocity space instead [2]. By considering all the
possible velocities attainable by the robot and applying those that lead away
from any impending collisions, the robot can navigate at fast speeds through its
environment.

Remark: This is useful when maps of some fixed obstacles (e.g. walls)
are known, but other dynamic obstacles (e.g. people) are unknown.

3.4. Rule Based Methods

Rule-based methods can be employed as an alternative to field-based
methods. These can be simple rules of the form developed by almost every
hobby robotics [5] - “if left sensor active, turn right”, or they can be generalized
by fuzzy logic and machine learning techniques.
 Remark: Simple rules are easy to use, but are often very limited to the
environments. Fuzzy control systems, employing fuzzy set theory, have been
proven to produce better performance than simple rules. However, these too are
limited to environments.

3.5. Rules Learning Techniques

Aoki used a robot programmed only with its basic set of performable
actions, and allowed the robot to learn its fuzzy rules by exploring its
environment, using a reinforcement learning algorithm [8]. Other rule learning
methods have included neural networks, as used by Tsoukalas and
evolutionary/genetic techniques, as used by Ram.

Remark: The robot develops the rule set that is suited to its environment
in a methodical way. However, these rules still need to be automatically
generated for each new environment. For generalized method, a great deal of
time needs to be spent in learning every possible environment the robot may
encounter.

4. Global Navigation
There are many kinds of global navigation systems. The most commonly
used methods are described here.

4.1. Search Trees

Common techniques for building search trees come from computer
science. These include brute-force search of the known map. This involves
constructing all the possible paths, and discarding those which are no optimal. A
dynamic search method developed by Stentz and Hebert [7] is an improvement
of the brute force algorithm. This was functionally similar, but claimed to be
over 200 times more efficient.

Remark: This efficiently solves the path planning problem, even in high
dimensions.
4.2. Global Planning in a Dynamic Environment
It is well-known because any algorithm to determine this will not
complete in reasonable time for real-time operation. Fujirriura developed a
“Time-minimal Path Planner”. In this method the robot first computes the
motion of every obstacle in its environment, then plans the longest distance it
can cover in the shortest amount of time. The robot moves along this time-
minimal path, and then plans the next path. Thus a sequence of motions leading
to reaching the goal is achieved.

Remark: The drawback of this method is that the motion of all obstacles
must be known in advance, making the system impractical for general use.

4.3. Probabilistic techniques

Cassandra Simmons and Howard work with policy-based techniques,
adapted from Markov’ decision processes using certain policy for combining the
desired action at all possible locations. The robot progresses forward by
iteratively acting on the combined action. After each iteration, the set of possible
locations decreases, until the robot reaches the goal. Takeda developed a
planning process which takes into account the likelihood of localization errors in
the development of paths [1].

Remark: Although this can technique achieve, all of above techniques

will break down if the location is known only probabilistically.

4.4. Hybrid-Integrated Approaches

In this approaches the whole motion problem can be considered as a
single task. This requires an integrated model of both goal-acquisitions as well
as collision avoidance. Xiao [4] describes an Evolutionary Planner/Navigators
(EP/N) which in parallel compute the optimal path using an evolutionary
(genetic) algorithm, while feeding in new obstacle information directly into the
search algorithm.
 Remark: While this work does not directly relate to the dynamic
problem, it showed considerable promise.

5. Scope and Challenges

This survey show that there is still scope for developing more efficient
motion planning algorithms with obstacles that will produce better quality routes
by addressing several challenging problems as follows:
(a) In a static or dynamic environment for successful navigation of a
mobile robot to the goal, two significant problems should be
resolved. The first one is real-time recognition of the obstacles. The
other is producing the shortest and safe path dynamically.
(b) Some of the above mentioned approaches used the kinematics
constrains of a mobile robot, while the other ignored it. It would be
considered a benefit for a specific navigation scheme to take some
of actual-world factors that affect the plan of mobile robot steering
schemes into account, because this makes it easier to transform into
a practical application.
(c) Another important factor which has a strong effect in dynamic
motion planning is the relative velocity. It is defined as the relative
velocity vector of the adjacent obstacle forward with regard to the
robot.
(d) Heuristic procedures do not guarantee the discovery of a solution.
Thus, to enhance the efficiency of these methods, one of the
excellent ways is to integrate some of them together.

Furthermore, multiple optimization goals, multiple robot direction,

uncertainty in detecting, forecast, path control, and many others constitute
numerous extra challenges in mobile robotics.
6. Conclusion
At the present time, improvement in mobile robot navigation has been
increasingly inspired by new applications in different sectors. This paper
discusses many methodologies used in mobile robot navigation. In spite of the
fact that several efficient techniques have been improved, the multitude of robot
navigation issues has been steadily growing. These problems involve
determination of collision-free route, shortest path, low runtime, modeling of
changing environment, multiple optimal functions and dynamic constraints.
These restrictions craft robot navigation issues to be more challenging and
require more strong and efficient algorithms.
REFERENCE
[1] A. Howard, “Probabilistic Navigation: Copying with uncertainity in
Robot Navigation Tasks”, PhD thesis. The University of Melbroune,
1999.
[2] Dieter Fox, Sebastian Thrun, Wolfram Burgard, “The Dynamic Window
Approach to Collision Avoidance”, IEEE Magazine on Robotics
&Automation, Vol. 4, No. 1, 1997.

[3] J. Borenstein and K. Yoram, “The Vector Field Histogram-Fast Obstacle

Avoidance for Mobile Robots”, IEEE Transactions on Robotics and
Automation, Vol. 7, No. 3: 278-288, 1991.

[4] J. Xiao, Z. Michalewicz, L. Zhang and K. Trojanowski, “Adaptive

evolutionary planner/navigator for mobile robots”, IEEE Transactions on
Evolutionary Computation, 1997.
[5] K. Ehjimura, “Motion Planning in Dynamic Environments”, Springer-
Verlag, 1991.
[6] K. Yorem and B. Johann, “Potential Field Methods and Their Inherent
limitations for Mobile Robot Navigation”, Proceedings of the 1991 IEEE
 International Conference on Robotics and Automation Sacramento.
California, 1991.

[7] Stentz A. and Hebert M., “A Complete Navigation System for Goal
Acquisition in Unknown Environments”, Intelligent Robots and Systems,
Proceedings, IEEE/RSJ International Conference on Human Robot
Interaction and Cooperative Robots, vol.1: 425– 432, 1995.
[8] T. Aoki, T. Suzuki, and S. Okuma, “Acquisition of optirnal action
selection in autonomous mobile robot using learning automats
(experimental evaluation)”, IEEE Conference on Fuzzy Logic and Neural
Networks on Evolutionary Computation, 1995.

Author:
Captain Zin Maung Cho, a graduate of B.C.Sc, was commissioned from
Defence Services Academy in 2007. He earned his M.Tech at Moscow State
University of Railway Engineering (Russian Federation) in 2012. Now he is
serving as an assistant lecture of the Department of Electronic, Communication
and Automation System, DSSTRC.