Simultaneous Localization and Mapping using a Differential Robot

OData support
Supervisor:
Kiss Domokos
Department of Automation and Applied Informatics

In today’s world it is easy to gain access to quality sensors and very high performance

computers and microcontrollers, thanks to recent advancements in technology. Reliable

and affordable robot hardware and software components are more widely available than

ever before. For these reasons, we have decided to develop a mobile robot that is able to

map and simulate its own environment, inside of this environment can determine its own

location, and is finally able to autonomously plan its own movement to arrive at certain

coordinates on the map.

To accomplish these goals we have developed a sensor system of our own, with the help

of which a map can be created in which the robot can perform its own localization. In

other words, we are replacing a costly LIDAR (laser scanner & rangefinder) sensor with

infrared distance sensors and servo motors. One of our main principles (and conditions) is

that all of our components throughout this development process must be easily obtainable.

To solve the problem of navigation we implement our own version of a Simultaneous

Localization and Mapping algorithm, published in scientific literature, and we use it to

process the data of the various sensors we employ. We use the EKF (Extended Kalman

Filter) variant of the widespread SLAM algorithm used in mobile robotics, which conceives

the core of the robot’s software. The input of the Kalman filter algorithm in our imple-

mentation consists of arrays of lines. The lines, in practice, represent the features of the

environment, and can be extracted from things such as walls and similar obstacles.

With the use of our sensor system, we gather point clouds and process them, using an

algorithm that was implemented by another student who also worked on the same robot,

to provide extracted line features for our SLAM algorithm. In addition to this, in order to

evaluate our SLAM method, we use the assistance of a path planning algorithm that was

also developed by this student.

The main goal of this study is to analyze these algorithms both in terms of simulation

and practical implementation. We perform the simulations using V-REP (Virtual Robot

Experimental Platform), in connection with our ROS nodes, and we employ a primitive

rover as our staring point, which from our perspective can be viewed as a partial solution for

our hardware problem. Our final objective is to demonstrate that a mobile robot equipped

with only simple sensors can be able to perform autonomous navigational functions as well.

Downloads

Please sign in to download the files of this thesis.